feasibility study for implementation of automotive ...857714/fulltext01.pdf · feasibility study...

BACHELOR’S THESIS Mechanical Engineering, Product development with design Department of Engineering Science

August 03, 2015

Feasibility study for implementation of automotive measuring method in aerospace industry Robin Söderblom Staffan Jonsson

BACHELOR’S THESIS

i

Feasibility study for implementation of automotive measuring method in aerospace industry

Summary

This thesis comprises an investigation in order to find possibilities to implement the method

used in the automotive industry to automatically generate a collision free measurement

program within the aircraft components manufacturer. The purpose with the study was to

compare and analyse the different methods used to generate measurement programs at GKN

Aerospace Engine Systems in Trollhättan, National Electric Vehicle Sweden (NEVS) and

Volvo Cars Corporations (VCC).

The study was conducted through meetings, observations and questionnaires with staff from

the geometry assurance engineering (GAE) departments and measurement departments in

each company. By mapping the virtual GAE process started from concept development in

CAD to the measurement phase in which components are measured in coordinated

measuring machines (CMM), a chain of activities was analysed.

NEVS and VCC are today using RD&T and IPS to generate optimized CMM programs in

which a time efficient measurement path can be generated. This method was compared with

the current approach at GKN Aerospace where they use one supplier for offline CMM

programming (OLP) software solutions and CMMs. They are thereby working in a closed

system where the OLP communicates with the CMM by supplier specific methods. The

automobile manufacturer NEVS and VCC, in contrast, uses a DMIS protocol which is an

ISO and ANSI standard.

The study shows that an implementation of the software used by the Swedish automobile

manufacture NEVS and VCC at GKN Aerospace in Trollhättan, may not have any

significant improvements regarding time savings and thereby no economic benefits.

However, the approach for generating an optimized measurement program in RD&T and

IPS may have major improvements in other facilities within the aerospace industry which

has also resulted in an instruction manual to be used for potential implementation.

Date: August 03, 2015 Author: Robin Söderblom, Staffan Jonsson Examiner: Mikael Eriksson Advisor: Timo Kero, Semcon Sweden AB Johan Lööf, GKN Aerospace

Anders Appelgren, University West Programme: Mechanical Engineering, Product Development with Design Main field of study: Mechanical Engineering Education level: first cycle Credits: 15 HE credits Keywords Measurement program, Measurement preparation, RD&T, IPS, Geometry

assurance Publisher: University West, Department of Engineering Science,

S-461 86 Trollhättan, SWEDEN Phone: + 46 520 22 30 00 Fax: + 46 520 22 32 99 Web: www.hv.se


ii

Acknowledgement

This project has been performed as a bachelor’s thesis, comprising 15 ECTS credits at

University West in Trollhättan. Before the period of this thesis a pre-study of 7,5 HE credits

was carried out by Robin Söderblom. The pre-study consisted of a literature study to obtain

knowledge that forms a theoretical framework for the geometry assurance process. The

content in chapter 2 and 3.1 in this thesis work overlap fully or partially with contents from

the pre-study; The Basics of Geometry Assurance, Product development PUC540, see

Appendix C. Except for this part the work during the thesis has been equally distributed.

Due to many practical moments during this project, the work has been dependent on the

contribution from many key participants. First of all we would like to thank our supervisors

Timo Kero, Team Manager for the group Geometry & Integration at Semcon Sweden AB,

Johan Lööf, Method Specialist at GKN Aerospace Sweden AB and Anders Appelgren,

Research Engineer at the department of Engineering Science at University West for their

strong support and commitment throughout the project.

We would also like to show our gratitude to Jie Shao, Sujith Guru and Jukka Pekka Mäki at

Semcon Sweden AB for their patience and helpfulness to provide knowledge and better

understanding of GD&T and RD&T. Furthermore, we would like to express our

gratefulness to Peter Josefsson and Maria Kvist, Measurement Specialists at NEVS, Roger

Andersson, Measurement Specialist at VCC as well as the Measurement Specialists Anders

Olausson and Sven-Olof Karlsson at GKN Aerospace, for your reception and the time you

reserved for us during the visits to your departments.

Next, we would like to thank Henrik Stranne at Hexagon Metrology for providing

information of the CMM equipment at Innovatum, Johan Torstensson at Fraunhofer

Chalmers Centre, for the kinematics in the CMM model and Svante Augustsson at PTC for

the useful information of the rapid prototyping machine as well as Hans Gustavsson at

Precuratum for having shared his knowledge and gave an introduction to the CMM. The

physical measurement test would not been possible without theirs involvements.

Additionally, we would like to thank Lars Lindkvist, Associate Professor at the department

of Product and Production Development at Chalmers University of Technology, for

accessing the RD&T software, and Tomas Hermansson at Fraunhofer Chalmers Centre, for

the access to the IPS software.

Gothenburg, June 2015

Robin Söderblom Staffan Jonsson


iii

Table of Contents

Summary .............................................................................................................................................. i

Acknowledgement ............................................................................................................................ ii

Symbols and Glossary ...................................................................................................................... v

1 Introduction ................................................................................................................................ 1 1.1 Company Description ..................................................................................................... 1 1.2 Background ....................................................................................................................... 2 1.3 Scope .................................................................................................................................. 2 1.4 Objectives and Limitations ............................................................................................. 2 1.5 Pre-Study ........................................................................................................................... 3

2 Geometry Assurance ................................................................................................................. 4 2.1 Concept Phase ................................................................................................................. 5

2.1.1 Locating Systems ................................................................................................ 5 2.1.2 Choosing Locating Scheme .............................................................................. 6 2.1.3 P-frame ................................................................................................................ 7 2.1.4 Stability Analysis ................................................................................................. 8 2.1.5 Statistical Variation Simulation ....................................................................... 11 2.1.6 Seam Variation Analysis .................................................................................. 14 2.1.7 Tolerance Allocation ........................................................................................ 14

2.2 Verification Phase .......................................................................................................... 15 2.2.1 Inspection Preparation .................................................................................... 15

2.3 Production Phase ........................................................................................................... 15

3 Equipment and Software Description .................................................................................. 17 3.1 RD&T .............................................................................................................................. 17

3.1.1 Stability Analysis ............................................................................................... 17 3.1.2 Statistical Variation Simulation ....................................................................... 17 3.1.3 Contribution Analysis ...................................................................................... 18 3.1.4 Inspection Preparation .................................................................................... 18 3.1.5 Offline Programming for CMM .................................................................... 18 3.1.6 Documentation ................................................................................................. 19

3.2 CMM ................................................................................................................................ 20 3.3 IPS .................................................................................................................................... 21

4 Methods for Situation Analysis .............................................................................................. 22 4.1 Meetings .......................................................................................................................... 22 4.2 Questionnaires ................................................................................................................ 23 4.3 Observations ................................................................................................................... 24

5 Development of the Instruction Manual .............................................................................. 25 5.1 Information Retrieval .................................................................................................... 25 5.2 Cad Modelling ................................................................................................................ 25 5.3 Tolerance Dimensioning ............................................................................................... 26 5.4 Prototyping ..................................................................................................................... 28 5.5 Measurement Preparation ............................................................................................. 28 5.6 CMM Inspection ............................................................................................................ 29

6 Results ........................................................................................................................................ 31 6.1 GKN Aerospace ............................................................................................................ 31 6.2 Volvo Cars ...................................................................................................................... 34 6.3 NEVS ............................................................................................................................... 36


iv

6.4 Instruction Manual ......................................................................................................... 39

7 Analysis ...................................................................................................................................... 40 7.1 Study ................................................................................................................................ 40 7.2 Instruction Manual ......................................................................................................... 43

8 Conclusions ............................................................................................................................... 44 8.1 Future Work ................................................................................................................... 44

References ........................................................................................................................................ 46

Appendices

A. 2D Drawings

B. Questionnaire

C. Pre-study: The Basics of Geometry Assurance

D. Instruction Manual


v

Symbols and Glossary

Δ Delta

3D Three dimensional

6σ Six Sigma

ANSI American National Standard Institute

CAD Computer Aided Design

CAT Computer Aided Tolerancing

CMM Coordinated Measuring Machine

CNC Computer Numerical Control

DCC Direct Computer-Control

DMIS Dimensional Measuring Interface Specification

DP Design Parameter

GAE Geometry Assurance Engineering

GD&T Geometric Dimensioning & Tolerancing

FR Functional Requirement

IPS Industrial Path Solutions, Software for path simulations

ISO International Organization for Standardization

JT A lightweight file format for industrial automation systems and integration.

Primary used in industrial cases to capture and reuse 3D product definition

data. JT is defined as ISO 14306.

KPS Control/Quality Process Management. A database that stores measurement

information.

Monte Carlo An algorithm used to simulate mathematical and physical systems.

MP Measuring point

NEVS National Electric Vehicle Sweden. Former SAAB

OEM Original Equipment Manufacturer

OLP Offline Programming

PLM Product Lifecycle Management

PMI Product and Manufacturing Information

Probe Sensor for geometrical inspection

R&D Research & Development


vi

RCA Root Cause Analysis

RD&T Robust Design & Tolerancing

RMS Root Mean Square

RPM Rapid Prototyping Machine. Adds material layer by layer to create a 3D-

structure

RSS Root Sum Square

Translator Software that reads native and DMIS CMM program languages

UG NX CAD software developed by Siemens

VCC Volvo Car Corporation

VRML Virtual Reality Modelling Language. A standard file format defined as

ISO/IEC 14772, which integrates 3D multimedia and graphics that can be

dynamically modified.


1

1 Introduction

The aerospace and automotive industry puts a lot of effort in product and production

development to be a competitive player on the global market. More focus on higher quality

and faster lead times at lower costs are fundamentals to keep up with the competition. The

principle of creating conditions for high quality products during early stages of the product

development has become more important. This since the costs of changes during later stages,

when specifications are set, may have a significant increase, see Figure 1. That is why the

geometrical variations during production are critical to quality characteristics, which has to

been taken into account.

This chapter describes the companies behind this project, the background to why this is

carried out as well as the scope and objectives of this thesis work.

Figure 1. The costs of changes during the development process may increase rapidly due to where they occur [1]. The relative costs-axis uses a logarithmic scale.

1.1 Company Description

This project is collaboration between GKN Aerospace Engine Systems in Trollhättan

Sweden and Semcon in Gothenburg, Sweden. These companies are described below.

GKN Aerospace is one of the world leader of supplying aerostructures, engine systems,

nacelles and transparencies to the aviation industry. They employ 12,000 people across four

continents and 35 facilities to support the military and civil markets with highly complex

metallic and composite assemblies for aerostructures and engine products [2] [3].

GKN Aerospace Engine Systems is a sub-division of GKN Aerospace and operates from

five facilities where the manufacturing plants are located in Sweden, Norway, Mexico and


2

USA. They have a development office in India and their head office in Trollhättan, Sweden.

GKN Aerospace Engine Systems in Trollhättan is divided into four business areas; Engine

products, Space, Military engines and Engine services [4] [5].

Semcon is an international consultant company with focus on engineering services and

product information. Today they have about 3,000 employees in ten countries around the

world, with their head office located in Gothenburg, Sweden. They offer services within the

areas of design, product and production development, project management and product

information [6] [7].

1.2 Background

As a manufacturing industry, GKN Aerospace strives for continuous improvements to

compete on the global market. They put a lot of effort in research and development (R&D)

to obtain market shares through advanced technology. The methodology of geometry

assurance, see chapter 2, are therefore implemented in the development process to assure

high quality through robust design. Within the methodology, several tools and methods are

used for measurement planning at GKN Aerospace. By investigating how the measurement

planning process at GKN Aerospace may be more time efficient than their current method,

the development process may receive shorter lead times and cost reductions, without quality

losses.

The result of the process improvement may be of interest to Semcon who strives to increase

their knowledge and experience within product and production development.

1.3 Scope

The purpose of this project is to investigate the feasibility for implementing the measurement

preparation method used by the automotive industry in the aerospace industry, and if such

implementation within all GKN Aerospace facilities may reduce costs and lead times.

1.4 Objectives and Limitations

The objective of this thesis work is to analyse and compile a comparison between the

different methods used to generate measuring programs at National Electric Vehicle Sweden

(NEVS), Volvo Car Corporation (VCC) and GKN Aerospace Engine Systems in

Trollhättan. The result should also lead to a written instruction manual where the proposed

activities to generate measurement programs in RD&T and IPS are described.

The instruction manual is limited to describe the general procedure to generate and simulate

a draft of a measurement program in RD&T and IPS, see Fel! Hittar inte referenskälla..


3

Figure 2. General process map for a virtual product development process started with functional requirements (FR) as input parameters. This report describes the development process from virtual geometric analyses to running physical measurement tests in CMM

1.5 Pre-Study

To understand the methodology and get an overall perspective of the geometry assurance

engineering (GAE) process, a pre-study was performed during the course; Product

Development II, PUC 540, at University West.

The literature study was compiled from scientific publications and PhD theses within the

field of the geometry assurance methodology. The general approach for improved quality

through robust geometry design is presented together with basic theories and examples.

The result is attached in Appendix C and parts of the content are presented in chapter 2 and

3.1.

Modelling using CAD

CMMMeasurementpreparation

2D and 3D drawings

Geometrical analyses, CAT

Pre-study Instruction Manual

Virtual Geometry Assurance Engineering Process

Scope of the report


4

2 Geometry Assurance

A common issue in manufacturing industries today is dimensional variation stack ups where

every single component have its own dimension variation that are included in an assembly.

This can either cause unexpected geometrical variation that propagates during the

production. This may lead to products that do not fulfil the aesthetical, functional or

assembly requirements. These geometrical quality problems contribute to high costs for

rework, market delays and bad publicity due to changes in product or production [8].

Geometry assurance is a methodology to manage variation and secure form, function and

assembly already in the concept phase. This is done by creating a robust design. A robust

product may be defined as a design that is insensitive against uncontrolled variation or

disturbance that may affect the performance, called noise factors, see Figure 3. These may

generally originate from manufacturing processes, temperature, wear, weather conditions and

so on. A good quality product may be characterized as a design that should be robust to

noise. The activity to improve the geometrical robustness of a product is called robust

geometry design [9].

Figure 3. A robust design is characterized by its insensitiveness to input variation (x) which affects the output characteristics (y) [9].

This chapter describes the basics of geometry assurance methodology and a set of activities

that supports the quality improvement process. To enable a high impact of the methodology,

the approach may begin in the early product concept phase where only a few design

characteristics have been developed. Virtual parts and subassembly models are used to

analyse the concept design parameters (DP) to be evaluated against the functional

requirements (FR), continuously throughout the concept, verification and production

phases, see Figure 4.


5

Figure 4. The approach of geometry assurance engineering.

2.1 Concept Phase

In robust geometry design the main source of variation occurs mainly during the

manufacturing process. By increasing the robustness of a design in early concept phase, wider

tolerances may be used on input parameters which may result in decreasing manufacturing

costs. The robustness of the concept design is therefore optimized by locating systems and

evaluated by stability analyses as early as possible. The aesthetical quality level of the concept

assembly is calculated and visualized by statistical variation simulations which are verified

against the assumed production systems.

2.1.1 Locating Systems

The main task within geometry assurance engineering (GAE) in the early concept phase is

to optimize the position of locators in a way to minimize the variation amplification, which

enables wider tolerances on input parameters. Thus, the robustness of the design increases

[9].

Locating schemes are used for positioning a part or sub-assembly in its correct position by

locking its six degrees of freedom in space during simulations, manufacturing, assembly and

inspections. Locating schemes uses locating points called locators which are strategically

placed on a part or sub-assembly. These theoretical locating points are realized by physical

planes, holes and slots. The locators are extremely important since fixture tool variation will

be transmitted into parts and subassemblies which contribute to the robustness of the design

concept [9]. The relation and dependencies can be formulated as equation (1).


6

[𝑟𝑜𝑏𝑢𝑠𝑡𝑛𝑒𝑠𝑠𝑣𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛

] = [𝑥 0𝑥 𝑥

] [𝑙𝑜𝑐𝑎𝑡𝑜𝑟𝑠

𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒𝑠] (1)

This equation indicates that the robustness is controlled only by the locators and should

therefore be focused on in early product and process concept phase. The main task in robust

geometry design is to place the locators in a way to minimize the variation amplification [9].

This locating procedure is supported by the stability analysis that will be described in chapter

2.1.4.

2.1.2 Choosing Locating Scheme

There are different types of locating schemes used in a variety of situations. The most used

locating schemes are presented below and other less frequently used are only mentioned.

3-2-1 locating scheme for rigid parts

Figure 5 illustrates an orthogonal 3-2-1 locating scheme with its six locating points.

There are three groups of locating points called primary, secondary and tertiary. These points

are described as follows:

- The primary locating points A1, A2 and A3, controls three degrees of freedom;

translated in Z (TZ) and rotation around X (RX) and Y (RY). These three points

define plane A.

- The secondary locating points, B1 and B2 control two degrees of freedom;

translation in X (TX) and rotation around Z (RZ). These two points define the

secondary locating plane B, perpendicular to A.

- The tertiary locating point C, control one degree of freedom; translation in Y (TY).

This point defines plane C, perpendicular to plane A and B.

In reality, the problem with all types of locating schemes is that they are coupled by nature.

This means that one locating point controls more than one degree of freedom. The 3-2-1-

system is the least coupled locating system since it enables the rotation and translation to be

minimized (decoupled). The ideal locating system is when one point only controls one degree

of freedom. Figure 6 shows the main couplings for the 3-2-1 locating scheme [10].

This system is the most commonly used and is easier to use and understand among the

locating systems. It can be applied on non-prismatic parts that are assumed to be rigid [10].


7

Figure 5. 3-2-1- locating scheme are the most frequently used

locating system [10].

A1 A2 A3 B1 B2 C1

TZ

RX

RY

TX

RZ

TY

Figure 6. The coupling model for 3-2-1 locating system.

The columns represents the points inserted and the rows

defines as translated or rotation vectors. The grey area

indicates the couplings [10].

Other locating systems

There are other variants of locating schemes used for different types of robustness analyses.

The 3-2-1 system is used for orthogonal rigid parts only. When working with non-rigid parts

which are allowed to deform or bend during positioning such as sheet metal or plastics,

clamping forces and part stiffness has to be involved to predict robustness and variation.

One example of locating system for non-rigid parts is the N-2-1 locating scheme, see [11].

For geometries with more irregular shapes there may not be possible to use locating schemes

with orthogonal localization directions. Here, the 6-points locating scheme may come in

handy with its 6-different directions. It is similar to the 3-2-1 system but allows the geometry

with non-orthogonal surfaces to be locked in its 6 degrees of freedom [12].

2.1.3 P-frame

When working with positioning systems the usual notation P-frame is used for locating

schemes. Every part have one local P-frame in general, often referred to as the master

location system that positions the part to the mating target P-frame on another component

or subassembly [13]. See Figure 7.


8

Figure 7. The 3-2-1 locating scheme referred to as the local P-frame positioned to the mating target P-frame [13].

2.1.4 Stability Analysis

The stability analysis is a tool to analyse the geometrical robustness by evaluating the coupling

amplification and how much variation introduced to the component caused by the locators

[14]. To get a fairly understanding of the evaluation method, the theory of axiomatic design

will be explained as well as the theory behind the stability calculations.

2.1.4.1 Axiomatic Design

Axiomatic design is defined as the mapping process between customer needs trans-formed

into functional requirements (FR), design parameters (DP) that physically satisfies the FRs,

and the process variables (aij) that represents the partial derivate aij=∂FRi/∂DPj at a specific

design point. The process variables are included in the design matrix [A] also called the

coupling matrix and may be written as equation (2) [13] [15].

[𝐴] = [

𝑎11 𝑎12 ⋯ 𝑎1𝑛

⋮ ⋱ ⋮𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛

] (2)

The design equation may be expressed as equation (3) [15].

{𝐹𝑅} = [𝐴]{𝐷𝑃} (3)

By using the design equation (3) in an example, the approach may be illustrated as equation

(4) [13].

[𝐹𝑅1

𝐹𝑅2

𝐹𝑅3

] = [𝑎11 0 00 𝑎22 00 0 𝑎33

] [𝐷𝑃1

𝐷𝑃2

𝐷𝑃3

] (4)

In equation (4), there are three function requirements (FR) that may be satisfied by three

design parameters (DP). The left side of the equation, FRs, may represent “what is wanted


9

in term of design goals” and the right side, [A] and DPs, represent “how we hope to satisfy

the FRs.” This design equation is the simplest case of design due to its non-diagonal elements

that are zero, a12= a13= a21= a23= a31= a32= 0. This is characterized as an uncoupled design

which means that one output parameter is controlled by only one input. The diagonal

characteristic is the most preferable design solution due to easier possibilities to change FRs

or DPs later in the product or production phase. This situation often occurs in parallel

assembly case where all parts are attached to “ground” by its own P-frame and has no

influence from other parts or P-frames. See Figure 8 [13].

In serial assembly cases, every part are attached to another part in a hierarchical order starting

with part A, controlled by its own P-frame as an example. The following attachments B, C,

D etcetera is controlled by its own P-frame and every P-frames mounted previously in the

assembly as illustrated in Figure 9. This is characteristic for a decoupled design and may be

written as equation (5) [13].

[𝐹𝑅1

𝐹𝑅2

𝐹𝑅3

] = [𝑎11 0 0𝑎21 𝑎22 0𝑎31 𝑎32 𝑎33

] [𝐷𝑃1

𝐷𝑃2

𝐷𝑃3

] (5)

A decoupled design is an acceptable design if performed in the correct order to prevent time

consuming tuning if necessary [13].

Figure 8. Parallel assembly model. Each P-frame

controlling its own P-frame only [13] .

Figure 9. Serial assembly model. The last P-frame (part

or subassembly) attached is controlled by all P-frames

[13].

A parallel assembly solution may therefore be preferably due to its less sensitivity to

adjustments during manufacture and assembly than serial assembly solutions [13].


10

2.1.4.2 Robustness Evaluation

The variation for a sub-assembly or assembly is calculated by varying each locating point, i

for a part with a small increment Δinput. Δoutput/Δinput can be determined in the X,Y and

Z directions for a number of n output points of the geometry. The root mean square (RMS)

values for all output points corresponding to variation in the locating points, are then

calculated. See equation (6)(7)(8) [9].

𝑅𝑀𝑆𝑥,𝑖 = √1

𝑛∑ [

(𝑥−𝑥𝑛𝑜𝑚)

Δinput]

2𝑛1 (6)

𝑅𝑀𝑆𝑦,𝑖 = √1

𝑛∑ [

(𝑦−𝑦𝑛𝑜𝑚)

Δinput]

2𝑛1 (7)

𝑅𝑀𝑆𝑧,𝑖 = √1

𝑛∑ [

(𝑧−𝑧𝑛𝑜𝑚)

Δinput]

2𝑛1 (8)

The resulted RMS values represent the mean influence of all locating points, i, which are

calculated in each direction separately to evaluate the total positioning evaluation goodness

for the total Root Sum Square (RSS) magnitude. The RSS influence in all locating points, i,

are calculated in each direction as well. See equation (9)(10)(11) [9].

𝑅𝑆𝑆𝑥 = √∑ 𝑅𝑀𝑆𝑥,𝑖26

𝑖=1 (9)

𝑅𝑆𝑆𝑦 = √∑ 𝑅𝑀𝑆𝑦,𝑖26

𝑖=1 (10)

𝑅𝑆𝑆𝑧 = √∑ 𝑅𝑀𝑆𝑧,𝑖26

𝑖=1 (11)

The RSS magnitude, equation (12), is used as a sensitivity value to evaluate how a certain P-

frame controls the stability of a certain part in a design [9].

𝑅𝑆𝑆𝑥,𝑦,𝑧 = √𝑅𝑀𝑆𝑥2 + 𝑅𝑀𝑆𝑦

2 + 𝑅𝑀𝑆𝑧2 (12)

To evaluate the coupling dependencies for an assembly, two measures are introduced as

reangularity (R) and semangularity (S). Read more about those in [15]. These values represent

the diagonality of the design matrix and are defined as equation (13) and (14) [9] [13].


11

𝑅 = ∏ [1 −(∑ 𝑎𝑘𝑖𝑎𝑘𝑗

𝑛𝑘=1 )2

(∑ 𝑎𝑘𝑖2𝑛

𝑘=1 )(∑ 𝑎𝑘𝑗2𝑛

𝑘=1 )]

1/2

𝑖=1,𝑛−1𝑗=1+𝑖,𝑛

(13)

𝑆 = ∏ [|𝑎𝑗𝑗|

(∑ 𝑎𝑘𝑗2𝑛

𝑘=1 )1/2]𝑛𝑗=1 (14)

aij is an element from the design matrix described earlier, and n is the number of rows in the

design equation. If R=S=1, the design matrix represents an uncoupled design which is the

theoretical ideal. R and S provide good possibilities to evaluate the degree of coupling in an

early concept assembly to avoid unnecessary costs for changes in production ramp-up.

Concepts requiring tight tolerances due to tolerance chains may be sorted out in early stage

using the stability analysis [13].

2.1.5 Statistical Variation Simulation

In order to control if the resultant variation for a part or assembly, caused by tolerances,

meets the FR´s, a tolerance analysis is commonly used. The purpose of using tolerance

analysis is to determine the effect of variations caused by each specified tolerance, called the

contributor. All the known tolerances that effects the total variation of a dimension

contributes to a tolerance chain, called stackup, see Figure 10. This analysis is thereby known

as the stackup analysis or design assurance [16].

Figure 10. Variation caused by a few individual tolerances in a stackup may result in a massive resultant [16].

A tolerance chain arises when a critical component dimension is dependent of another

individual dimension. Figure 11 illustrates a simple one-dimensional tolerance chain

occurring on a vehicle floor consisting of a tunnel for the cardan shaft and two separate

floors panels on both sides. The resulting width of the floor is controlled by the sum of every

component dimension. The overall variation is thereby controlled by every individual part.

This makes the design very sensitive and results in difficulties to adjust the floor width during

assembly. By using this solution, the manufacturing process may be more expensive to assure

high quality due to tighter tolerances for every component [17].

Figure 11. The total floor width is controlled by each individual part dimension [17].


12

The most efficient solution from an economic perspective is to eliminate all stackus by

redesign the concept. Figure 12 illustrates a different solution to the stackup problem. By

allow the parts overlapping each other and including a fixture to adjust the floor width during

assembly, the solution become more robust and minimizes the high demands on tight

tolerances for every part. The fixtures on both sides control the overall width and the fixture

pin controls the tunnel position which may be adjusted just before production starts. These

fixtures may be used in the assembly process and are being removed later [17].

Figure 12. Components are overlapping to avoid stackup dependencies and fixtures are used to control the total floor width [17].

If robust design and ease of adjustments to a specific quality level during production, shall

be obtained, tolerance chains must be avoided and this is one example to an alternative

solution to eliminate a critical tolerance chain.

2.1.5.1 Stack-up Models

When working with complex three-dimensional (3D) assembly design, tolerance chains may

be difficult to identify and handle in production. Tolerance analysis is therefore preferably

performed before the final geometry is set to detect potential tolerance stackups and increase

geometrical robustness.

There have been abundance methods for performing tolerance analysis for rigid components

developed through the years. The worst-case (WC) model, Monte Carlo and a number of

statistical models presented in Figure 13 are variants used for the analysis [18].

Figure 13. Different mathematical models for tolerance analysis presented in Chase [18]. Mean shift and Six

Sigma are variants of Root Sum Square.


13

The most widely used statistical models is the Root Sum Square (RSS) which will be described

and compared with two other frequently used models, namely the worst-case and the Monte

Carlo model [18].

Worst Case

The worst-case model is the simplest model among the others presented in Figure 13 and is

based on the arithmetic law. It assumes that all tolerances in a stackup are at its extreme limit

simultaneously to obtain the worst possible combination of parts. If the WC stays within the

required tolerance limits, there are no rejected assemblies needed. The stack formula is non-

statistical and may be written as equation (15) [8] [18].

𝑑𝑈 = ∑|𝑇𝑖| ≤ 𝑇𝐴𝑆𝑀 (15)

Where dU is the predicted assembly variation, Ti is the component tolerance allowed for one

specific dimension and TASM is the maximum respectively minimum tolerance limit required

for the overall assembly.

This method is commonly used by designers in early concept phase to assure that their

assemblies stays within the specified tolerance limits, but is also preferably performed during

manufacturing where the production volume is low and the tolerance chain is short [8].

Statistical RSS

By adding variations to the calculation, the predicted limits are more reasonable due to its

statistical probabilities of the possible combinations. The RSS model is the most simple of

the statistical models and assumes a normal distribution of component tolerances in an

assembly. This model is applicable in high production volume and longer tolerance chains

but may have an optimistic result. The predicted assembly variation may be written as

equation (16) [8] [18].

𝑑𝑈 = √∑ 𝑇𝑖 2 ≤ 𝑇𝐴𝑆𝑀 (16)

The equations presented for WC and RSS only illustrates the basic theory for the models and

these become more complex in real assembly systems. Chase and Parkinson [19] writes about

factors such as sensitivities, correction factors and mean shift factors that are introduced in

the calculation to get a more realistic prediction.

Monte Carlo Samples

Monte Carlo simulations have been a very frequently used tool for tolerance analyses since

it handles both non-normal distributions and nonlinear assembly functions [8]. It uses a

number generator to randomly simulate the effect of variation from manufacturing


14

processes. By generating output parameters for every dimension of an assembly and

iteratively continues until a sufficient number of iterations have been simulated, a percentage

of rejected assemblies are estimated based on the specified tolerances.

This method is sample-based but the simulation becomes statistical with enough number of

samples and requires more than 100 times the CPU capacity compared to WC and RSS [19].

The Monte Carlo simulation is very useful in 3D tolerance analyses and is used as basis for

the majority of today´s computer aided tolerancing (CAT) systems [8].

2.1.6 Seam Variation Analysis

The relation between two parts over a specified distance in assembled products describes the

most frequently used quality characteristics for geometrical variation evaluation. The quality

of the split-line between two body panels of a vehicle, for example the doors, hoods and

panels are critical quality characteristics due to its functional and esthetical aspects. The door

must be possible to open without any conflict with surrounding parts while the split-line,

mainly translated by the gap, flush and parallelism, must satisfy its desired quality

requirements [20].

The gap represents the distance between two parts in a common plane, See Figure 14, while

the flush refers to the distance between two parts perpendicular to a surface or a plane. These

dimensions are often measured or calculated for two specific points, where each point is

located on each part in a specified 2D-plane [14].

Figure 14. Seam variation illustrated by gap and flush [14].

2.1.7 Tolerance Allocation

Optimizing performance, quality and production costs often requires tolerance allocation to

strategic allocate the critical contributors. Tolerance allocation is often described as the

opposite of tolerance analysis due to the purpose to break down tolerance chains in an

assembly, to locate the individual contributors [8]. By performing a contribution analysis, the

contribution for each part variation is calculated and the most critical tolerances may be

prioritized and investigated. The basic formula for each part variation may be written as

equation (17) for worst case scenario and equation (18) for the statistical root sum square

[18].


15

𝑊𝐶: %𝐶𝑜𝑛𝑡 = 100𝑇𝑖

𝑇𝐴𝑆𝑀 (17)

𝑅𝑆𝑆: %𝐶𝑜𝑛𝑡 = 100𝑇𝑖

2

𝑇𝐴𝑆𝑀2 (18)

2.2 Verification Phase

In the verification and pre-production phase the virtual product models are physically

realized to be tested and verified against the production system. Here, adjustments are made

for both the product and the production system to prepare for full production volume. In

geometry design the verification phase involves inspection preparation where inspection

routines and strategies are established. The virtual assembly model is used to minimize the

geometry errors by locator adjustments and to support the inspection plan [8].

2.2.1 Inspection Preparation

The aim of inspection preparation is to find the optimum set of inspection points that

captures product information to verify if adjustment, correction or compensation is

necessary. Often the number of inspection points becomes quite large in pre-production to

access a large quantity of process information to monitor the actual process level. During

full production the numbers of inspection points are reduced to only capture key dimensions

on every product and the measuring process is carrying out in every assembly stages to

monitor the quality level through the entire assembly process [8].

2.3 Production Phase

During the production phase all adjustments from the pre-production are completed. The

product geometry design satisfies the FR´s, the geometrical tolerances may be accepted and

the product is in full production. Samples may be analysed during the production process

for distinguishing between the common causes due to noise factors and assignable causes

often due to defect raw material, operator errors, machine and tooling errors. These

variations may be controlled and minimized or eliminated by Root Cause Analyses (RCA) or

by the Six sigma (6σ) approach [8].

Figure 1 introduced the importance of designing a robust product in early stage due to more

expensive changes in production. Part variation arises in the manufacturing process and

amplifies due to wear in manufacturing tools over time. Together with fixture and assembly

variations, the geometrical variation of the final product is given as a result [12]. See Figure

15.


16

Figure 15. Cause and effect diagram of assignable causes that may contribute to the overall assembly variation [9].

TotalVariation

Assembly Variation

PartVariation

Process Variation

MachinePrecision

Process Variation

AssemblyPrecision

Design Concept

Robustness

Assembly Process

Manufacturing Variation


17

3 Equipment and Software Description

A few equipment and software’s have been used during this project. The CAD software, NX

Unigraphics from Siemens [21], was used to create virtually assembly models of a test product

which then was analysed and prepared for measurement in RD&T (Robust Design &

Tolerancing) and IPS. The final measurement program was thereafter verified and tested

during measurement inspection in a CMM.

3.1 RD&T

RD&T is a math based software tool for statistical variation simulation that accounts for

geometrical variations during manufacturing and assembly. The tool was developed by

RD&T Technologies and the software allows products to be simulated and visualised in early

concept phase, long before any physical prototypes are being produced. Various concepts

may thereby be analysed and compared to improve the quality of decision making.

This tool was initially developed to evaluate variation within mass production for automotive

and aerospace industries. Today, it is supporting OEMs, suppliers and consultants for

various applications as an aid for geometric quality improvements [22] [23].

The software can be customized with variety of modules to support all phases in the GAE

process. From early concept phase where only a shell model exists, to the production. These

modules supports as a toolbox for virtual geometry assurance and some of the tools used in

this project are presented below.


The stability analysis evaluates the geometrical robustness of a component or assembly in

early concept phase. By introducing small variation increments (Δ input) in X, Y and Z

direction for each locating points, Δ input/Δ output may be determined, see Figure 3. The

software then calculates the sum of variation in each point and the RSS value for all points

in a system may be defined and the robustness of the design can be evaluated. See equation

(9), (10) and (11). RD&T also shows which areas having the highest output amplification by

colour coding the absolute magnitude or the robustness may be analysed in each X, Y and Z

direction separately to the improve the decision making [14].

To improve the robustness of the component or assembly, the locators may be repositioned,

since the robustness is only controlled by its locating points, see equation (1).


To be able to determine the quality level for a product before production realization, the

geometrical variation for critical features must be predicted. By performing a statistical

variation simulation on critical assembly dimensions, the product may be analysed and

improved before the first prototype is build. By using the Monte Carlo method, the software

randomly generates all input parameters within the defined distributions for every part and

simulates the resulting output parameters. For a number of iterations the simulation predicts


18

the expected output standard deviation, range, mean value and capability indices etcetera, for

the specified dimension. Thereby can the critical feature be determined and evaluated against

the requirements [8].

3.1.3 Contribution Analysis

When all manufacturing data, i.e. tolerances and distributions, is defined for a system, the

focus is to optimizing the tolerances until they satisfies the FR´s, manufacture and costs

constraints. RD&T provides the ability to trouble shoot and optimizing the tolerances by

listing all part tolerances that contributes to the total variation within a system. The

contribution analysis function in the software presents a ranked list that reflects the overall

influence of position, variation range and variation direction for each point. This enables the

opportunity to improving the design robustness as well as reducing costs for unnecessarily

tightened tolerances, since only the major contributors are modified [24].


The module for inspection preparation in RD&T assists the verification process by optimize

and specify the inspection points [8]. These points represent the target location in space

where the measuring equipment should measure.

3.1.5 Offline Programming for CMM

RD&T provides the possibility to generate measurement programs for CMMs. The working

procedure consists of three stages; defining features, designing of the measuring program

and visual simulation. A brief introduction of each step is given below.

Define feature

The inspection points, also called measuring points (MP), defined earlier in the process, only

specifies where the information of the geometry should be gathered. If the measuring

machine uses a touch-probe, see chapter 3.3, and the measuring point represents a point in

the centre of a hole or a slot, there is no contact. The measurement equipment is thereby

unable to find the geometry.

The solution is to define an inspection feature that consists of two or more MPs. These MPs

is the definition of a touching point where the probe is in contact with the target geometry.

By defining an inspection feature, several MPs may be translated into a measurement vector,

which in the case of an hole feature, may represent a hole centre. See Figure 16.


19

Figure 16. An inspection feature is created in RD&T to be able to measure the hole centre. It may be defined by seven MPs, where three MPs represent a plane, which defines the normal, and four MPs that defines the inner circle. Those MPs creates a vector in the hole centre which represents the measure point.

Designing the measurement program

When the inspection points and features are defined, a new measurement program can be

created in RD&T. Raw DMIS code, see chapter 3.2, and additional information needed for

the measurement can thereby be written manually.

The final measurement program may thereafter be visually simulated corresponding the real

measurement scenario.

3.1.6 Documentation

RD&T supports the R&D process with all engineering documentation. Measurement

drawings, product requirement drawings and reports can be created, since the requirements

are already set earlier in the GD&T process. The purpose of the documentation is to define

all the functional and aesthetical requirements of the component or assembly. This in order

to make it possible for the manufacturer to fulfil the specified requirements.


20

3.2 CMM

Coordinated measuring machines (CMM) are used to physically measure the actual geometry

of an object. The recorded result may then be compared with the nominal shape and

dimensions as might be specified in a 3D model with product and manufacturing information

(PMI) or on a part drawing. These CMMs can be of different types, either manually or direct

computer-controlled (DCC). Manually driven CMMs controls the probe position by human

operator movements while the measurements are provided by digital outputs. The operator

may then record the result either by hand, paper printouts or by computer-assisted data

recordings. Direct computer-controlled CMMs on the other hand, operates like computer

numerical controlled (CNC) machine tools. The movements are motorized in the orthogonal

X-, Y- and Z-directions and are controlled by computer programs [25]. These orthogonal

motions may be performed by various physical configurations. Figure 17 shows the most

common types of CMMs where the probe is mounted on a moving bridge structure. Each

axis is moved relative to a fixed table on which the measured object is positioned.

The measurement may be performed by several different methods, either by touch-trigger

probes or noncontact methods. In the letter, methods like photogrammetry, white light

scanner, laser scanner or laser interferometer systems are methods for optical measurements

[26]. Electrical field, radiation or ultrasonic techniques are other noncontact methods which

are non-optical [25]. The most common measurement technology today is using touch-

trigger probes which make contact with the part surface. The computer is thereafter

recording the coordinated position of the probe immediately after contact [25].

DCC CMMs may be controlled by offline programs (OLP). The program are therefore

prepared off-line based on measurement requirements from part drawings long before the

execution. This allows the programming to be accomplished in parallel with other

measurement preparations for the same CMM. The preparations may also be carried out

while the machine is running another program [25].

The majority of off-line programs for CMMs are today based on CAD systems in which the

geometrical data is used. The link between the CAD and the CMM systems are

communicated via protocols. Dimensional measuring interface specification (DMIS) is an

ISO and ANSI standard for two-way communication of inspection data which allows CAD

system to communicate with any CMM [27] [25].


21

Figure 17. The moving bridge design for CMMs are the most frequently used in industry today [25].

3.3 IPS

IPS, Industrial Path Solutions, is a math based tool for generation of collision free assembly

path, mainly used for off-line programming of robots and CMMs [28]. The software is

developed by Fraunhofer Chalmers Centre [28] to help simulation engineers to optimizing

the route from one point to another by taking into account all surrounding geometries at the

workstation. By import a scene geometry consisting of the CMM, the objects, fixtures and

surroundings, as VRML or JT-files from any CAD system, IPS will find an efficient path

[29].

This tool has been used to optimize assembly path for both manually and automated use, as

well as finding the most time efficient route to carry out a series of measurements for

coordinated measuring machines [29].


22

4 Methods for Situation Analysis

GKN Aerospace asked for an investigation in which the possibility of implementing similar

measurement preparation process as the automobile industry is using today. The Swedish

automobile manufacturer VCC and NEVS have set high demands on aesthetics and fit on

their products for decades. Hence, these are of interest to GKN Aerospace to be compared

with.

When performing an investigation, reliability and validity must be taken into consideration.

Validity is the value of the data collected which measures how well it corresponds to the

reality and the reliability reflects the degree of how the information is true [30]. The methods

for obtaining new data were thereby based on primary data. Primary data collection is

considered to be more reliable then secondary data, since the secondary data only relies on

information that has gone through several individuals [30].

The research should also take into consideration whether inquiries are structured or

unstructured. Structured approaches forms the investigation process; sample, design and

objectives, where the answers from the respondents are predetermined. This is classified as

a quantitative research. Unstructured approaches allow, by contrast, flexibility in the

respondent’s answers, which may account for different opinions about an issue or

description of an observed situation. This approach is thereby classified as a qualitative

research [31].

This investigation was carried out by mapping the virtual GAE process for each company.

The research was conducted by three different methods for qualitative data collection;

meetings, observations and questionnaires, since operator opinions and time constraints was

taken into consideration.

4.1 Meetings

This project has consisted of several meetings where a lot of information have been obtained.

There are different types of meetings; information meetings, working meetings, decision

meeting, negotiating sessions and review meetings [32]. This study was primary conducted

by information meetings where information exchanged with the affected persons.

There is of importance to prepare each meeting by sending an agenda with a clear purpose

to every participant. This may contribute to an efficient and rewarding meeting since the

participants have a chance to prepare and contribute while the meeting is carried out. If there

is a lack of communicated objectives, the meeting could be unstructured and time consuming

which may result in high costs without reaching results. Bo Tonnquist suggests to almost

preparing every meeting as thoroughly as a court hearing, to make sure the time is spent

efficiently [32]


23

Prior to each meeting, the following should be of consideration [32].

Have a clear purpose

Send an invitation and agenda at least one week before the meeting

Documentation - invitation, agenda, protocol and decision making

Hold of schedule

Manning right people

The information meetings at GKN Aerospace, NEVS and VCC were implemented in

conjunction with pre-prepared interview questions. The objective was to gather information

about the overall GAE process of each company with focus on their current measurement

preparation process.

4.2 Questionnaires

A questionnaire consists of a list of questions addressed to many recipients, which in turn

will respond, either on written paper, internet or by other form [30]. The major difference

between a questionnaire and an interview is that in the latter it is the interviewer who poses

the questions that may observe and record while the respondents reply. Whilst in the former,

the respondents are recording the answers themselves [31].

When choosing a questionnaire or interview schedule, there may be of importance to

consider the advantages and disadvantages of the two methods. The decision may be based

on the following criteria [31].

The nature of the investigation - If the survey asks for sensitive or personal

opinions. The respondent may therefore request anonymity.

The geographical distribution or limitation of the study population - There

may be geographical, time or costs limitation if the respondents are scattered over a

wide area.

The wording and format of questionnaires are out of importance as they affect the

respondent’s willingness to give a good answer. The questions should thereby be relevant

and appropriate. There are two ways the questionnaire may be formulated; closed-ended or

open-ended. In the former the possible answers are predefined and the respondent selects

the answers that describe the respondent´s answer best. Open-ended on the other way, gives

the respondent the opportunity to answer in words, since the possible response are not given

[31]. This opens up for open-minded answers and explanations, but may also be time

consuming for both the respondent and the investigator [30].

Questionnaires have been used during this project to answer questions that were forgotten

or unanswered during the meetings. The choice of using this data collection method was

made due to time constraints. To be covering the whole GAE process of each company,

these where formulated to gather detailed information about their work process, their


24

activities and the order they were executed in. Questions about disadvantages of their current

process as well as improvements were included. The questions were formed as open-ended

internet mails.

4.3 Observations

Observational research is a method for gathering primary data by observing relevant

situations, people and actions [33]. The advantage of an observational research is the

systematic and selective way of adopting information by watching and listening to a

phenomenon or interaction. There are a variety of situations in which the observational

research may come in handy. This data collection method may be suitable when; for example,

ascertain or study the function performed by an operator, study personality behaviour,

learning purposes or when accurate information cannot be induced by questioning. Other

methods for primary data collection may not even be appropriate when the subject is

involved in the interaction, that they may not be able to contribute to objective materials.

This is where observation may be the advisable approach to gather the required information

[31].

In general, there are two types of observation; participant observation and non-participant

observation. The former describes a situation when the researcher is participating during

observational activities in the same manner as the group member. Thus they are not aware

that they are being observed. A non-participant observation, in contrast, is when the

researcher observes the activities by watching and listening without getting involved in the

situation. The researcher will thereby remain as a passive observer which may draw

conclusions from a distance [31].

A non-participant observation was carried out at the measurement department of GKN

Aerospace in Trollhättan. The objective was to gather information about the use and

interaction with their software for measurement preparation. Since the measurement

specialists have been working with the same interface for a long time, they may have

difficulties to express or find disadvantages as well as advantages with their solutions. New

detailed information could be highlighted by observing the working procedures while the

involved specialist prepared a product for the measurement.


25

5 Development of the Instruction Manual

To compile an instruction manual for the measurement preparation process with the use of

RD&T and IPS, several activities was carried out. New knowledge was therefore obtained

by studying the GAE process for geometric dimensioning and tolerancing (GD&T), RD&T

and IPS. A test model was then created in CAD and thereafter realized to be physically

verified in a CMM. The following sections describe these activities in more detail.

5.1 Information Retrieval

To be able to create the instruction manual a lot of information had to be gathered as well

as obtaining new knowledge. A pre-study was carried out before this project to obtain

knowledge within the field of the geometry assurance methodology, see Appendix C.

Thereafter, internal GD&T material from Semcon [34] and coarse literature from

Precuratum [35] was studied as well as ISO 1101:2013 [36] and 5459:2011 [37].

5.2 Cad Modelling

An example model was created to illustrate a common case within an assembly situation.

There may be different solutions to achieve a robust assembly design. An early design of

phone components was therefore created. The design illustrates three electrical circuit boards

(ECB) with different solutions for attachments on a mobile base board (MBB), see Figure

18 below. Even the fixture that was needed for the inspection was modelled, see Figure 19.

The difference between the three ECBs is the dimensions, shapes and position of the

positioning holes. All ECBs have the same purpose which is to be attached to the MBB. The

main attention where focused on aesthetical requirements which corresponds by the gap

between the ECB and MBB when they are assembled. It is also of consideration to counteract

flush between the top surfaces of the two components. The different hole features results in

different robustness for the assembly in which will affect the quality.


26

Figure 18. Visualisation of the 3D models; Mobile Base Board, at the left, and three different designs of Electrical Circuit Boards.

Figure 19. Visualisation of the fixture.

5.3 Tolerance Dimensioning

When the design in early phase had been defined for the MBB and ECB and the gap and

flush requirements had been specified. The work started in RD&T. All three parts were

imported as VRML files with information about the geometry of the parts. Thereafter the

fixture was defined as a rigid part. The MBB was positioned to the target fixture with 6-

points locating system which in this case corresponds to the 3-2-1 system. The first tree

points define the bottom surface of the MBB and prevent translation in Z-direction and

rotation around X- and Y-axis. The second two points, B1 and B2, prevents translation in x-

direction and rotation around z-axis. The last point, C1, prevents translation in y-direction.

See Figure 20 below.


27

Figure 20. 6-points locating systems is used to locate the components in space.

Thereafter the ECB was positioned to the MBB by using 6-points locating system which, in

this case, represents the location for the suggested pins and holes. A stability analysis was

performed on the first concept design. This design was thereafter improved in five steps until

the final design was specified. See Figure 21 for the robustness improvements.

Figure 21. Results presented in color-coding from stability analyses on the first design (to the left) and the fifth design (to the right).

Due to the design changes, the 3D model where updated and approximated tolerances were

applied. These tolerances were defined in RD&T where a statistical variation simulation of

10.000 iterations was carried out. The result showed whether the applied allowed variation

fulfilled the requirements. 2D-drawings of the MBB and ECB parts with GD&T can be seen

in Appendix A.


28

5.4 Prototyping

Prototypes of the MBB and ECB was created in physical shapes by a 3D rapid printer. Totally

four unique components were produced. One MBB and one specimen for each ECB design,

see Figure 22. The realization was carried out by a Wanhao duplicator 4, desktop 3D printer

at Innovatum in Trollhättan. The models where build with 0,2 millimetre layers of PLA

plastic with 10 percent infill ratio.

Figure 22. The result from 3D printout of the MBB and ECB with different hole features.

5.5 Measurement Preparation

The measurement preparation for inspection with CMM has been conducted in off-line

mode with the use of RD&T and IPS. This process is divided into three software activities;

RD&T to IPS and back to RD&T. The process to generate an inspection program for the

CMM starts in RD&T. The first step was to import all parts or an assembly in VRML-format

and define the inspection and reference points for the parts or assembly. When this was

conducted the inspection direction for the probe to approach each specific measuring point

was specified by creating features. A feature may be defined as; circle, cylinder, edge point,

plane, point, slot, sphere, stud and tetragon. Each feature has individually defined approaches

that determine how many number of points that are needed. A plane needs, as an example,

at least three points to be defined. Further on a program was created where all features were

added and a program header with additional information was inserted. In this stage, the

current measuring points was exported as a .pdmi-file into IPS.


29

Next step in the preparation process was to import the CMM geometry and the assembled

measuring object with its corresponding fixture. The CMM was imported in IPS-format since

kinematical information has to be included. The measuring objects were imported in VRML-

format and the program from previously step with all measuring features. The CMM

geometry and the measuring object were thereafter positioned to correspond to the reality.

Afterwards IPS identifies all possible ways to measure each measuring point. The software

then automatically optimizes the inspection plan, in which order an time efficient collision

free path is created. This inspection plan is thereafter visually simulated and exported as a

.pdmo-file into RD&T.

Back in RD&T, the inspection plan from IPS was imported and added to the measurement

program. The program was thereby exported as a DMIS file. This preparation process is

further described in more detail in Appendix D.

5.6 CMM Inspection

CMM inspection has been performed at the Production Technical Centre (PTC) in

Trollhättan. Equipment used for the inspection is a DEA Global Advantage 7.10.7 CMM

with a TESASTAR-m probe head. The probe used was a Renishaw TP200 with 1 millimetre

stylus attached. See Figure 17. This CMM was DCC operated by PC-DMIS software.

The inspection started with calibration of the above mentioned probe system, the method

and equipment used is a calibration ball with a fixture placed on the CMM table. The

procedure is to first select the approach angels that were going to be used during the

inspection. This was examined in PC-DMIS. The stylus ball was thereafter manually placed

above the calibration ball and moved slowly downwards to get contact. Afterwards the CMM

knows where the calibration ball is located in space and DCC calibration was initiated to

automatically calibrate the probe system for the predefined approach angels.

When the calibration had been performed the fixture with the MBB part were positioned on

the measurement table to create a first alignment of the reference system. The approach was

to manually measure four points on the fixture plate to create a XY-plane. Further on the

second alignment was created corresponding the 2D-drawings, see Appendix A. A plane and

two points was therefore used to specify the position and thereby the reference system of

the MBB part. This information was manually added to the imported DMIS code that was

created in previous chapter 5.5. When all preparation was conducted the physical

measurement was performed fully DCC by the DMIS code. Figure 23 illustrates the gap and

flush features defined in RD&T which letter was evaluated during the measurement.


30

Figure 23. Visualization of features used for inspection corresponding the gap and flush between MBB and ECB. The flush features have been simplified in the illustration, with a smaller caption area then in the measurement program. Flush (squares) where evaluated in Z-direction and gap (circles) in X- and Y-direction.


31

6 Results

This chapter presents the results from the study and the development of the instruction

manual. Following a comparison between two automobile manufacturing companies, VCC

and NEVS, will be compared with GKN Aerospace method to generate CMM programs.

6.1 GKN Aerospace

The virtual GAE process at GKN Aerospace Engine Systems in Trollhättan has been studied

through meetings, questionnaires and observations. In a meeting with the method specialists

Johan Lööf and Anders Per Johansson at the GAE department, they described their current

approach and the history of GAE within the company. The results are presented in the

following.

GKN Aerospace in Trollhättan has been working with one dimensional WC calculations for

a long time since the former Volvo Aero. Statistical RSS calculations were only carried out

when they have sufficient knowledge of the tolerance chain. The use of WC calculations are

very common due to the nature of the tolerance-stack-up´s. The GAE department has

thereafter been established in recently and is today supporting the development and

production process with the use of RD&T. They are therefore able to do advanced statistical

3D tolerance analyses to improve the decision making. The involvement of GAE in their

product development projects within the company have varied over time.

The current process for measurement preparation and CMM measurement within the

company was described in a meeting with the method technician Anders Olausson at the

measurement department. The following information was obtained during this meeting.

The measurement preparation process begins with an incoming order to the measurement

department. A new component will thereby be prepared for a physical measurement. The

newly received order is then scheduled. Depending on which components that are

prioritized, whether it is a prototype or a component in production, the order is selected and

the measurement preparation may begin.

The measurement department is today using CALYPSO which is a common system for

online and offline measurement programming. The software is supplied by Carl Zeiss

Industrial Metrology which also delivers the measurement machines used within the

department [38]. CALYPSO enables the measurement process to be completed within one

software suite. The software then communicates the measurement plan to corresponding

CALYPSO software which in turn operates Zeiss CMMs [39].

An offline programming session in CALYPSO was studied by a non-participant observation.

The approach for measurement preparation within the software was analysed by observing

the procedure for preparing an engine component. The observation was conducted with one

measurement technician. New information that was obtained during the session is presented

below.


32

CALYPSO has a feature-oriented interface which is user friendly and may be suitable

and easy to understand for non-skilled measurement technician.

The measurement path is manually defined by the measurement technician. The path

is thereafter simulated. If there is any collision with surrounding geometries during

the simulation, the software displays an alert. The path must then be manually moved

in the program.

There is no possibility to change the raw code manually.

CALYPSO enables good possibilities to perform the measurement with rotary tables

attached. By allowing the object to rotate, the measuring probe may then be placed

in a certain position and measure thousands of measurement points in a few seconds.

There is therefore no difference in time whether the CMM is capturing ten or

thousands of measuring points.

The measurement technician was pointed out during the observation that CALYPSO´s

simulation engine have not been able to correspond the reality at 100 percent. The simulation

has been improved since recently.

In a conversation with the measurement specialist Sven-Olof Karlsson at the measurement

department, he concluded that 90 percent of their measuring points are measured by rotating

the measuring object. Since the majority of their engine parts are axisymmetric they using a

rotary table during the measurement.

The approach to assure that all requirements are met for each manufactured component at

GKN Aerospace in Trollhättan has been analysed. By mapping their virtual GAE process,

the information flow is investigated. Figure 24 describes the company´s virtual GAE process

from the concept phase where the first concepts are designed, to the physical measurement

and outcome analyses. This process flow has been carried out through meetings and

questionnaires with the previous mentioned technician and specialists. The process is

described in the following.

GKN Aerospace in Trollhättan uses Siemens CAD software NX Unigraphics to design new

products. The virtual GAE process begins therefore with an early concept 3D model where

a few design features are defined. Each version of CAD models are then exported as a

Unigraphics part file (.prt) to Teamcenter which is their common product lifecycle

management-system (PLM). This PLM system is supplied by Siemens and is used to store

and manage product information for downstream applications throughout the development

and manufacturing process [40]. The 3D model is transferred to an internal storage within

the GAE department. The early design characteristics are imported as JT or VRML files into

the CAT software which then is being analysed. Stability analyses are performed in RD&T

and improvement proposals for increased robustness is thereafter communicated to the

design engineers. When the final geometry is completed and a robust design is obtain, the

design engineer creates requirement drawings. These tolerances are then evaluated in RD&T

where statistical variation simulations are performed. The result is thereafter communicated


33

back to the design engineers who may update the drawings. This is the current work flow in

those projects where the GAE department is involved.

Further on while the component is realized in the manufacturing plant, the measurement

department receives the order for measurement preparation. The measurement technician

then imports the CAD file from Teamcenter into CALYPSO where a measurement program

is carried out offline. All measuring points are specified by the measurement technicians who

interpret the information printed on the GD&T drawings. In those cases when there is time,

the technician may optimize the program manually to be more time efficient during the

physical measurement. The final measurement program is then exported to Teamcenter as

an internal file format for CALYPSO, see ① in Figure 24.

When the component or assembly has been manufactured and assembled, they are delivered

to the measurement department. Measurement operators mount the measuring object onto

a fixture which is positioned in the CMM. The measurement program is then imported from

Teamcenter into CALYPSO, which this time operates their Zeiss machine. The coordinate

system in the program must correspond with the physical measuring object which

theoretically is positioned in an unspecified space. The operator operates therefore the CMM

manually by a controller to capture five to six points on the object surface to define where it

is located. The measurement program is thereafter tested through a test run to clarify if the

alignment was correct and if the program corresponds to the reality. Subsequently, the

program is executed.

The result from the measurement is exported as an internal CALYPSO output file to

Teamcenter, see ② Figure 24. Selected data are thereafter imported to their analysis database

KPS. This is a non-visual analysis tool where quality managers may analyse the real output

values compared with the nominal. To analyse the data stored in KPS, material number and

requirement numbers has to be known. The GAE department are therefore looking for other

solutions for data analysis databases.

The answers from the questionnaire sent to Anders Olausson is attached in Appendix B.


34

Figure 24. The virtual GAE process flow at GKN Aerospace Engine Systems in Trollhättan.

Due to the size of aircraft components manufactured at GKN Aerospace in Trollhättan, the

measurement phase during the manufacturing process may result in a bottleneck. Table 1 is

a guideline to get a fairly understanding for the time required for measurement preparation

and execution. This guideline was conducted in a conversation with the measurement

technician Anders Olausson at the measurement department.

Table 1. Time required for different activities during measurement preparation and measurement.

Stage in process Activity Time

1 Measurement preparation (OLP) 1 month

2 Setup - Mounting and fixing 1 hour

3 Online alignment and verification 2 days

4 Physical measurement with CMM 1-10 hours

6.2 Volvo Cars

An information flow diagram and approach for the virtual GAE process at VCC has been

established. In a meeting with the geometry assurance engineer Roger Andersson at the

department for GAE body at VCC, he mediated their current approach for measurement

preparation and their exchange to RD&T and IPS.

VCC have been working with Dassault Systems Catia CAD software for a long time. Back

in 2012 they were preparing measurement programs in Audi AG´s Audimess [41] which was

an OLP software used within Catia V4 for generating DMIS programs. When VCC then

migrated to Catia V5, could Audimess thereby no longer be used since it was only supported

by the Catia V4. The automobile manufacturer was thereafter looking for the possibility to

find new suppliers of OLP solutions to generate DMIS programs. The developer of RD&T

CAD ANALYSISCMMMEASUREMENT PREPARATION

CAT

NX RD&T CALYPSO CALYPSO KPS3D

MO

DEL

(.p

rt)

2D D

RA

WIN

GS

3DM

OD

EL (

.jt/

.wrl

)2D

DR

AW

ING

S

3DM

OD

EL (

.prt

)2D

DR

AW

ING

S

MEA

SUR

EMEN

TP

RO

GR

AM①

MEA

SUR

EMEN

TP

RO

GR

AM①

MEA

SUR

EMEN

TR

ESU

LT ②

MEA

SUR

EMEN

TR

ESU

LT ②

WORK FLOW DATA FLOW IMPROVEMEN PROPOSALS

INTERNAL STORAGE

TEAMCENTER


35

was, at this time, working on a measurement preparation function which was of interest to

VCC. A development project in cooperation between VCC, FCC Chalmers and RD&T

Technology was initiated to develop a solution which today has become a way to

automatically generate a collision free measurement path for CMMs with the possibility for

time optimization. The first version of the measurement preparation function with the use

of RD&T and IPS was tested in early 2012 at VCC.

Their current approach by using RD&T and IPS during their GAE process has been mapped,

see Figure 25. The information flow was studied since VCC is using RD&T in several

activities throughout the process. The virtual GAE process may begin with an early concept

design which is virtually created as a 3D model in Catia V5. Each version of the model is

thereafter exported to Teamcenter. These Catia files are automatically converted into JT-files

on a daily basis in Teamcenter. They are thereafter imported into RD&T where the

robustness of the design is evaluated using the tool for stability analysis. Design changes are

then communicated back to the design engineers. Once the tolerances are set, they are

analysed in RD&T and the geometry assurance engineer supports the decision making during

the 2D drawing establishment.

The working procedure at the GAE department for measurement preparation begins by

importing JT-files from Teamcenter. The geometry assurance engineer then defines target

locations on the geometry were critical information must be gathered. Measurement points

are therefore specified based on experience. Measurement drawings are thereafter generated

in RD&T which judicial specifies measuring points in which external supplier should be

measure their products. These drawing are not needed for in house manufacturing. Further

on, measurement features are defined, whether there is a hole, slot or edge etcetera. that is

being measured. When all features needed for the measurement program are defined, a list

of all measuring points is exported as .pdmi into IPS.

CMM geometry including kinematical configurations and measuring object geometry with

corresponding fixture are here imported into IPS as well as the measuring points. All

geometries are thereafter positioned corresponding the real scenario during the letter

measurement. IPS then calculates all possible measuring positions based on reachability,

probe configurations and general predefined measurement settings. The software then

creates an optimized collision free measurement path. The level of optimization is based on

computer capacity and time required for the calculation. The result is thereafter visually

simulated corresponding the real measurement scenario in the CMM. The measurement plan

is then exported back into RD&T as .pdmo.

The .pdmo file is thereafter added to a measurement program in RD&T in which a DMIS

program is generated. When the physical product is prepared at the measurement

department, the measurement program is imported from Teamcenter into one of the

company´s CMMs. VCC have invested in CMMs over time and is therefore using different

CMM suppliers. Today they use solutions from Hexagon Metrology, LK and Metrologic.

Each supplier has their own program translator for their CMM, see ③ Figure 25. VCC is


36

therefore exporting their measurement programs in DMIS format in which all CMM

translators may adapt. Afterwards, a DMIS output (.dms) is created with information

concerning the actual geometry gathered from the measurement. This information is

thereafter sent to the analysis database CM4D where measurement technicians may analyse

the actual outcome compared to the nominal 3D model as well as the requirements.

Figure 25. The virtual GAE process flow at Volvo Cars in Gothenburg.

6.3 NEVS

National Electric Vehicle Sweden (NEVS), former SAAB, has been working with the GAE

approach in over 30 years and the methodology is highly integrated within the company. In

a meeting with the leader of geometry systems Peter Josefsson and Maria Kvist, head of

dimensional management, they described their current GAE process within the company.

The information gathered is presented in the following.

NEVS have been working with RD&T and IPS since 2014. They are therefore still

implementing the new approach in their current GAE process which is presented in Figure

27. Their virtual GAE process begins with 3D modelling of concept parts in UG NX. In this

stage it is just a simplified early stage design. The 3D model is then exported into Teamcenter

in as NX (.prt) and JT (.jt) file format. The JT geometry file is then imported into RD&T to

be improved the robustness of the design. These improvement proposals are thereafter

communicated back to the design engineers who may change the design in NX. The CAT

phase during the process may here be compared with VCC´s approach described in previous

chapter. 2D drawings are thereafter created based on the result from statistical variation

analyses in RD&T.


CAT

CATIA RD&T RD&T/IPS ③ CM4D

TEAMCENTER

3DM

OD

EL (

.CA

TPar

t)

3DM

OD

EL (

.jt )

3DM

OD

EL (

.jt)

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MP

-DR

AW

ING

S

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)

RD&T IPS RD&T.jt.pdmi (.pdmo)

INTERNAL PROCESS


CM4D

3D M

OD

EL (

.jt)


37

When GD&T requirements are set, may the measurement preparation activity begin. The

3D model is imported from Teamcenter into RD&T as VRML or JT format. The following

OLP process is equivalent with the approach at VCC. The internal activity begins with

defining the measuring points followed by specified features. Figure 26 presents the interface

in RD&T when specifying measuring points on a body sheet. The company have developed

a measurement strategy which defines what information should be gathered whether it is

features such as holes, slots or edge etc. The measurement strategy includes a naming

standard to facilitate the handling of the points. Thereafter a program is created where the

features is added. Measuring points and the object geometry are thereafter exported as .pdmi

file into IPS. The internal procedure in IPS is to position the 3D model of the CMM, fixture

and product in space, then decide which feature to be measured in the specified program.

IPS optimizes thereafter an collision free path for the CMM. The result from IPS is exported

as a .pdmo file back to RD&T where a DMIS program is created. The measurement program

may here be visually simulated in either IPS or RD&T.

Figure 26. Body sheet and fixture geometries are used for OLP in RD&T at NEVS. Measuring points are here specified in which geometry data can be captured in letter measurement. (Source: NEVS)

An internal file sharing system is used to export the measurement program into the CMMs

at NEVS. Since the automobile manufacturer uses different CMM suppliers, they have

various program translators for each CMM that reads the DMIS program while post

processing in the background, see ③ Figure 27. The result from the CMM inspection is

exported as a DMIS output file (.dms) into product quality validation software titled CM4D.


38

This software was developed in a cooperation between the automotive manufacturer SAAB

and DMC in year 2000. CM4D is referred to as DIDAS at NEVS and lists all components

for an assembly in a tree system. VRML and JT files are imported to visualise where in space

the measuring points are located as the outcome values from the physical measurement are

analysed against the nominal values. The variation simulation previously performed in

RD&T may here be compared to the actual outcome from the production. Peter Josefsson,

leader of geometry systems, mentioned in the meeting that the interface in CM4D is user

friendly even for non-skilled measurement technicians. Figure 28 gives a hint of the interface

in CM4D.

Figure 27. The virtual GAE process flow at NEVS in Trollhättan.


CAT

NX RD&T RD&T/IPS ③ CM4D

TEAMCENTER

3DM

OD

EL (

.prt

& .j

t)

3DM

OD

EL (

.wrl

& .j

t)

GD

&T

ITEM

S (P

MI)

3DM

OD

EL (

.wrl

& .j

t)

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)

RD&T IPS RD&T.wrl.pdmi (.pdmo)

3D M

OD

EL (

.wrl

& .j

t)FILESHARE CM4D


INTERNAL PROCESS


39

Figure 28. User interface in CM4D at NEVS. Inspection data from physical measurement from CMM are here compared with nominal values. The 3D model of a boot lid is used to visualise the location of measuring points. (Source: NEVS)

6.4 Instruction Manual

The instruction manual is describing the general process to generate CMM programs within

RD&T and IPS. The framework of the manual is based on informative screen pictures from

RD&T and IPS with additional text to each picture. The MBB and ECB parts have been

added to visualize the preparation process to make it easier to understand. Red marks are

here used to highlight the actual buttons described. The instruction manual is attached in

Appendix D.


40

7 Analysis

This chapter presents analysis and reflections regarding the results and methods used during

the study and development of the instruction manual. The analyses are based on the

presented results and reflections are based on the methods that have been used during this

project.

7.1 Study

The result presented in this report is mainly based on information gathered during

information meetings at GKN Aerospace, NEVS and VCC. All participants work within

relevant GAE areas in each company. GKN Aerospace´s measurement preparation software

CALYPSO has been briefly studied by a non-participant observation with one measurement

technician. The user approach may be personal which in turn may affect the observational

result. A questionnaire was also used to gather additional information and to understand

their GAE process. RD&T and IPS has been further investigated through educational

materials in which corresponds with the current approach for measurement preparation at

NEVS and VCC. The information retrievals resulted in a virtual information flow in which

has been validated by phone and email contact with GAE staff within each company.

GKN Aerospace in Trollhättan has chosen to only use ZEISS as supplier of measurement

solutions for measurement preparation and CMMs. They are thereby able to use a closed

system that communicates in between the processes. The observational research shows that

CALYPSO has a feature-oriented interface which compared to RD&T makes the measuring

point definition much faster. By clicking on feature geometry and defining a plane, several

measuring points may be defined in seconds. Measuring point definition in RD&T on the

other hand, is specified one by one, since the OLP function in RD&T is developed for free

form geometries as an automobile body for instance. RD&T does therefore not solve

measuring point definition for axisymmetric geometries as user friendly and time efficient as

CALYPSO. Due to measurement preparation times which may take up to month for one

component at GKN Aerospace, are the offline programming activity the most critical to

avoid bottlenecks in the daily production. There are thereby of importance to use OLP

software which enables fast measuring point definitions.

CALYPSO has not been able to simulate the measurement path corresponding the real

scenario at 100%. This may contribute to longer test runs of the program before actual

measurement during CMM inspection at GKN Aerospace. IPS on the other hand, simulates

the actual measurement scenario precisely, according to NEVS.

DMIS are widely used among today’s manufacturing industries. Since the protocol is set as

standard by ISO and ANSI there are advantages such as supplier independent multi-way

communication between different systems and CMMs. DMIS programs may also be created

manually as well as using software for automatic program generation. User training

requirements are therefore reduced since there may only be one program language to use.


41

GKN Aerospace is thereby lack of flexibility since they using CALYPSO´s closed

communication system which requires another ZEISS translator to read the program. They

are therefore dependent on one supplier of measurement solutions.

The automobile manufacturer NEVS and VCC are using RD&T and IPS for measurement

preparation. The majority of measuring points on body panels captures information from

freeform surfaces. These components are in many cases fixed in large complex fixtures in

which there is a need for time optimized measurement programs. RD&T and IPS was

therefore developed to automatically generate an optimized collision free measurement path

in which are programmed in DMIS. For measurement objects with hundreds of measuring

points, this may save time at both OLP and CMM measurement. There are also possibilities

to do a cluster analysis in RD&T, which is excluded from this report. The cluster analysis

may be used during the inspection preparation in which the majority of measuring points

can be reduced. The tool clusters several measuring points into one point which captures the

most critical dimensions [8].

The purpose of mapping the virtual GAE process in each company was to investigate the

information flow to find advantages and disadvantages with each process. Since the

automobile manufacturer uses RD&T in several activities and departments, there may be

information that can be transferred which in turn may prevent unnecessary rework in letter

processes. Information from the variation analyses in RD&T does never reach the letter

GAE or measurement department at NEVS and VCC today. Their current process does not

share information between the departments even though they use the same software for

robust design and measurement preparation. As previously mentioned, is the measuring

point definition a critical time consuming activity which affects the process flow in daily

manufacturing. Figure 29 illustrates an improvement proposal that has been conducted in

which information of points and attributes may be shared from the variation simulation

phase into the measurement preparation phase. This might thereby result in better

communications between the departments and a red thread through the process. NEVS have

already plans to investigate this possibility in the future. The geometry assurance engineer

Roger Andersson at VCC concludes that there are different models used in the variation

simulation phase and measurement preparation phase. The final model may not even be

analyzed in RD&T during the concept phase. It is therefore imported into RD&T first during

the measurement preparation phase. The points defined in the concept phase may be most

likely different to the measuring points and features which should define the actual

measurement geometry.


42

Figure 29. Proposed approach to share additional information into the measurement department by the use of RD&T.

Disadvantages with the current preparation process at NEVS and VCC was discovered to

be difficulties to use and navigate through RD&T and IPS. Since there are different mouse

commands for navigation in each software. Figure 25 and Figure 27 illustrates all files that

are imported and exported in between the software´s. This may also be ineffective since the

preparation require more than one software to be completed. CALYPSO may here be lack

of the functionalities that IPS provides. However, CALYPSO provides all the work to be

performed into one software suit. This solution may therefore be more beneficial for better

user experience.

In a meeting with the GAE department at GKN Aerospace in Trollhättan, they revealed that

their database for measurement analyses, KPS, is difficult to use since there is no visual model

to compare with. To find specific data material- and requirement numbers has to be known.

They are therefore searching for better alternatives for data analysis databases. The Swedish

automobile manufacturers are today using the interactive measurement analysis database

CM4D. All measuring data are here visually presented in the nominal 3D model. The tool

delivers ease of use and lists complete assemblies in a tree system in which the user may find

and analyze each components outcome individually.

GKN Aerospace´s current method for measurement preparation have been compared with

the method used by the Swedish automobile manufacturer. Offline programming in RD&T

and IPS results in a generated collision free measurement path in which is optimized for time

efficient measurement in any CMM. This study was thereby limited to investigate one

method for measurement preparation compared to the current process. There are several

other software´s available for offline programming of CMMs. Siemens NX provides

possibilities for automated offline programming of CMMs in associativity between NX CAD

and NX CAM [42]. ZEISS have developed the OLP software Caligo [43] which aimes at the


CAT

NX RD&T RD&T/IPS ③ CM4D

TEAMCENTER

3DM

OD

EL (

.prt

& .w

rl)

3DM

OD

EL (

.wrl

& .i

gs)

3D M

OD

EL (

.rd

t)

3DM

OD

EL (

.rd

t)

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MEA

SUR

EMEN

TP

RO

GR

AM

(.d

mi)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)

MEA

SUR

EMEN

TR

ESU

LT (

.dm

s)3D

MO

DEL

(.j

t & .w

rl)

RD&T IPS RD&T.wrl.pdmi .pdmo

INTERNAL PROCESS

WORK FLOW DATA FLOW PROPOSED WORK FLOW


43

automotive industry with freeform surface measurements and iDA in which generates DMIS

programs [44]. EMS PC-DMIS is standard equipment in many CMM brands which also

comprises solutions for OLP [45]. WENZEL have been developed the OLP Quartis which

is designed using Microsoft Office Fluent user interface [46] and the Metrologic is delivering

their OLP solutions in Metrolog XG [47].

7.2 Instruction Manual

Information gathered from the meetings at VCC and NEVS have built the framework for

the instruction manual; The Process to Generate CMM Program in RD&T and IPS. The

instruction manual had later on been tested and verified in a pilot case, where the MBB and

ECB part had acted as the product that where prepared for CMM inspection. This inspection

had not been possible to carry out without expert help from Johan Torstensson at FCC

Chalmers who had contributed with kinematic information of the specific CMM.

Measurement technician Hans Gustavsson at Precuratum had contributed with his time and

knowledge during the CMM inspection.

Because of the relatively small number of information sources it is reasonable to be critical

to the process presented in the instruction manual. This is something that may need to be

tested and verified by competent staff within measurement planning and geometry assurance

departments. During the CMM inspection at PTC the DMIS code where tested with a DEA

Global Advantage 7.10.7 CMM. The result from this test showed that the DMIS code could

be run by the CMM with the PC-DMIS software. But it required some changes because of a

default setting in RD&T where the probe was searching for a default reference system. Due

to this manual programming where performed in online mode. This is a weakness in the

development of the instruction manual. The manual must therefore be tested again and on

several different objects.

The instruction manual are only describing an general process to generate CMM programs

with RD&T and IPS, the product that is acting as an example are the MBB and ECB part.

The three different designs of the ECB parts are visualizing how important the locating

system is for a robust design. The design with the small holes was impossible to assemble

without finishing work of the envelope surface inside the holes. Which will lead to longer

lead times during manufacturing and increased manufacturing costs? The design with the

large holes will be possible to mount, but this is not robust and will failure to meet the

aesthetical requirements. The third design with the slot is the most robust design that was

produced, but this design could probably be improved by moving the slot from the lower

right corner to the upper right corner. This will increase the distance from the lower left hole

that is the first connection between the parts. Due to this the robustness will probably

increase but this need to be confirmed by a stability analysis in RD&T.


44

8 Conclusions

GKN Aerospace Engine Systems in Trollhättan is looking for alternative solutions for offline

programming of their CMMs in which may result in economic benefits. The investigation of

their current process for measurement preparation in comparison with the use of RD&T

and IPS shows that there may not be any significant improvements if implementing the letter

OLP solution. GKN Aerospace in Trollhättan is today using OLP and CMM solutions from

one supplier in which can deliver optimized information communication between the

measurement program and the CMM. An implementation of DMIS protocol

communication will thereby result in functionality losses since ZEISS have their own

solutions for OLP and CMM operations.

The study also shows that the automobile manufacturer NEVS and VCC have difficulties to

find a red thread though the GAE department and measurement departments in which more

information could be shared in RD&T. The automotive and aircraft engine component

manufacturer are both industries in which puts high demands on shape and fit. Even though,

there are differences in the components that are being controlled by CMMs. The automotive

industries measures mostly irregular freeform surfaces where aesthetical and functional

aspects such as gap and flush are critical to quality characteristics. GKN Aerospace in

Trollhättan, in contrast, manufactures up to 90 percent axisymmetric components in which

they uses rotary tables during in their CMM measurements. The solution allows the

measuring probe to be positioned in a specific position in which the rotary table rotating the

measuring object. Thousands of measuring points will thereby be measured in second. This

approach cannot be more optimized since there is no path to move through. However, there

may be advantageously to use RD&T and IPS to optimize the measurement path for the ten

percent which require irregular surface measurement methods.

The offline programming function in RD&T was developed to be implemented within the

automotive manufacturing applications. The functionality was first tested in 2012 which

means that the solution is still under development. RD&T Technologies has developed

flexible software which is adaptable for many calculation applications. GKN Aerospace has

great potentials to develop the OLP software with focus on round table functionality in

cooperation with RD&T Technologies in which may be adapted for the aerospace industry

applications with more advantages as result.

8.1 Future Work

This project has been resulted in a feasibility study for implementation of automotive

measuring method in aerospace industry. The investigation was conducted by a qualitative

comparison between to methods used for measurement preparations. The result previously

concluded presents that there are different advantages and disadvantages by implementing

RD&T and IPS into GKN Aerospace´s measurement process. There is of interest to

measure the total time required from OLP to performed CMM measurement compared to

CALYPSO and RD&T. The comparison should thereby be examined by offline


45

programming the same component, multiple times and by several operators to gather

quantitative data.

There are other facilities within GKN Aerospace where they uses older measurement

methods to control the quality of freeform surfaces. There may be of interest to apply the

study presented in this paper to investigate possible improvements in other measurement

planning processes as well as introducing the use of RD&T and IPS in other aerospace

activities.

The Instruction manual also needs to be tested and validated by more experienced staff

within RD&T and IPS.


46

References

[1] B. B. a. B. Klefsjö, Quality: from customer needs to customer satisfaction, Lund: Studentlitteratur, 2010.

[2] “About us,” GKN Aerospace, [Online]. Available: http://www.gkn.com/aerospace/aboutus/Pages/default.aspx. [Accessed 29 05 2015].

[3] “What we do,” GKN Aerospace, [Online]. Available: http://www.gkn.com/aerospace/aboutus/Pages/what-we-do.aspx. [Accessed 29 05 2015].

[4] “GKN Aerospace Sweden,” GKN Aerospace, [Online]. Available: http://www.gkn.com/aboutus/Pages/locations/gkn-aerospace-Trollhattan.aspx. [Accessed 29 05 2015].

[5] “GKN Aerospace Engine Systems,” [Online]. Available: http://www.gkn.com/aboutus/Documents/locations/trollhattan/booklet_engine_systems_final_screen.pdf. [Accessed 29 05 2015].

[6] “This is Semcon,” Semcon Sweden AB, [Online]. Available: http://www.semcon.com/en/About/. [Accessed 29 05 2015].

[7] “Services,” Semcon Sweden AB, [Online]. Available: http://www.semcon.com/en/Services/. [Accessed 29 05 2015].

[8] L. L. J. C. R. Söderberg, “Virtual Geometry Assurance for Effective Product Realization,” in 1st Nordic Conference on Product Lifecycle Management - NordPLM´06, Gothenburg, 2006.

[9] L. L. S. D. R. Söderberg, “Computer-aided robustness analysis for complient assemblies,” Journal of Engineering Design, vol. 17, no. 5, pp. 411-428, October 2006.

[10] L. L. J. S. C. R. Söderberg, “Managing physical dependencies through location system design,” Journal of Engineering Design, pp. 325-346, 2006.

[11] H. S. J. Y. J. X. Cai W, “Deformable Sheet Metal Fixturing: Principles, Algorithms, and Simulations,” J. Manuf. Sci. Eng, pp. 118, 318-324, 1996.

[12] S. Lorin, Geometric Variation Simulation for the Development of Products with Plastic Components, Gothenburg: Chalmers University of Technology, 2012.

[13] L. L. R. Söderberg, “Computer Aided Assembly Robustness Evaluation,” Journal of Engineering Design, pp. 165-181, 1999.

[14] L. L. R. Söderberg, “Stability and seam variation analysis for automotive body design,” Journal of Engineering Design, pp. 173-187, 2002.

[15] S. N. P., The principle of design, New York: Oxford University Press, 1990.

[16] O. Wagersten, “Visualizing the Effect of Geometrical Variation on Perceived Quality in Early Phases,” Chalmers University of Technology, Gothenburg, 2013.

[17] H. J. R. Söderberg, “Tolerance Chain Detection by Geometrical Constraint Based Coupling Analysis,” Journal of Engineering Design, pp. 5-24, 1999.

[18] C. K., Basic tools for tolerance analysis of mechanical assemblies, Utah: Brigham Young University, 2004.


47

[19] A. R. P. K. W. Chase, “A survey of research in the application of tolerance analysis to thedesign of mechanical assemblies,” Research in Engineering Design, vol. 3, no. 1, pp. 23-37, 1991.

[20] R. S. L. L. A. Dagman, “Split-line design for given geometry and location schemes,” Journal of Engineering Design, pp. 373-388, 2007.

[21] Siemens, “NX,” Siemens Product Lifecycle Management Software Inc., [Online]. Available: http://www.plm.automation.siemens.com/en_us/products/nx/. [Accessed 03 08 2015].

[22] T. K. e. al., “Virtual Variation Simulation Of CAD/CAM Template-guided Surgeries Performed on Human Cadavers: Part II,” The Journal of Prosthetic Dentistry, 2010.

[23] “The tool RD&T,” RD&T Technology, [Online]. Available: http://rdnt.se/tool.html. [Accessed 29 05 2015].

[24] Robust evaluation and tolerance analysis, Software Manual for RD&T, RD&T Technology, 2014.

[25] M. P. Groover, Automation, Production Systems, and Computer-Integrated Manufacturing, Harlow: Pearson Education Limited, 2014.

[26] M. P. P. J. Lööf, “Geometrisäkring inom ramen för L-FAM II,” Högskolan i Halmstad, Halmstad, 2007.

[27] “DMIS,” Inspec Softwaare Corp., [Online]. Available: http://www.dmis.com. [Accessed 29 05 2015].

[28] “Industrial Path Solutions,” FCC Chalmers, [Online]. Available: http://www.fcc.chalmers.se/software/ips. [Accessed 29 05 2015].

[29] “IPS Path Planner,” FCC Chalmers, [Online]. Available: http://www.fcc.chalmers.se/software/ips/ips-path-planner/. [Accessed 29 05 2015].

[30] J. A. Kylén, Att få svar, Stockholm: Bonnier Utbildning, 2004.

[31] R. Kumar, Research Methodology, California: Thousands Oaks , 2005.

[32] B. Tonnquist, Projektledning, Stockholm: Sanoma utbildning, 2014.

[33] P. K. G. Armstrong, Marketing: An Introduction, Upper Saddle River: Pearson Prentice Hall, 2009.

[34] Form- och lägetoleranser - Internal material, Semcon, 2011.

[35] Referenssystem och mätning i CMM, Internal material, Broddetorp: Precuratum.

[36] Svensk Standard SS-EN ISO 1101:2013, Swedish Standard Institute, 2013.

[37] Svansk Standard SS-EN ISO 5459:2011, Swedish Standard Institute, 2011.

[38] “About us,” Carl Zeiss Industrial Metrology, [Online]. Available: http://www.zeiss.com/industrial-metrology/en_de/about-us/welcome.html. [Accessed 29 05 2015].

[39] “Calypso Software from Carl Zeiss,” Inspection Engineering, [Online]. Available: http://www.inspectionengineering.com/Zeiss_Software.htm. [Accessed 29 05 2015].

[40] “Document & Content Management,” Siemens Industry Software AB, [Online]. Available: http://www.plm.automation.siemens.com/se_se/products/teamcenter/document-content-management/index.shtml. [Accessed 29 05 2015].


48

[41] J. M. G. Reinhart, “3D-Simulation: Schneller, sicherer und kostengünstiger zum Ziel,” Institute für Werkzeugmaschinen und Betriebswissenschaften, München, 1995.

[42] “Cmm Programming,” Siemens Product Lifecycle Management Software Inc., [Online]. Available: http://www.plm.automation.siemens.com/en_us/products/nx/for-manufacturing/cmm-programming/#lightview-close. [Accessed 29 05 2015].

[43] “CALIGO,” Zeiss Sweden, [Online]. Available: http://www.zeiss.se/industrial-metrology/sv_se/produkter/software/caligo.html. [Accessed 29 05 2015].

[44] “iDA,” Zeiss Sweden, [Online]. Available: http://www.zeiss.se/industrial-metrology/sv_se/produkter/software/ida.html. [Accessed 29 05 2015].

[45] “PC-DMIS CMM,” Hexagon Metrology, [Online]. Available: http://www.hexagonmetrology.se/PC-DMIS-CMM_388.htm#.VWc_zU-qpBc]. . [Accessed 29 05 2015].

[46] “Quartis,” Wenzel , [Online]. Available: http://wenzelamerica.com/products/software/quartis. [Accessed 29 05 2015].

[47] “Metrolog XG,” Metrologic Group, [Online]. Available: http://www.metrologic.fr/en-us/products/software/metrologxg.aspx. [Accessed 29 05 2015].

[48] “Main steps,” RD&T Technology, [Online]. Available: http://rdnt.se/steps.html. [Accessed 29 05 2015].


Appendix A:1

A. 2D Drawings


Appendix A:2


Appendix A:3


Appendix A:4


Appendix B:1

B. Questionnaire

Questionnarie sent to measuement technician Anders Olausson at GKN Aerospace

Questions Original Answers Translated Answers

Can you describe the process for measurement preparation at your department? Include the following into your answer.

What information do you receive into your department, i.e. the process inputs?

Innan vi kan starta programmering behöver vi detta (detta bör komma flera veckor innan detaljen kommer, beror på omfattning. En stor beredning tar ca 160h); Ritning med kravnummer, Kontrollplan/lista, eller markerad ritning, Information om övriga utvärderingar och processmått, CAD-modell i PRT-format.

Before we start the programming, we need the following (This has to arrive weeks before the component arrives, may depend on the extent. A big preparation takes about 160h); Drawing and requirement number, inspection plan/list or marked drawing, information about other evaluations and process indicators, CAD model in PRT format.

If you have time, feel free to explain your process flow for each activity. What takes place in parallel?

Programmeringen sker offline och parallellt med övriga metoder. Det som måste ske innan är indatan ovan. Därefter testkörs programmet, för att sedan frisläppas tillsammans med operationsunderlaget.

The programming occurs offline and in parallel with other methods. The indata described earlier must been received. The program is thereafter tested in a test run to be released together with the operations substrate.

How is the distribution of existing production and products in development undergoing measurement preparation?

70% är utvecklingsprojekt, 25% är förbättringar på befintliga produkter, 5% är forskning eller utveckling i tidigt skede (prototyper, enstaka detaljer).

70% is in development projects, 25% improvement of existing products and 5% R&D in early phase (prototypes and single parts).

What/which operations takes place in CALYPSO? Please explain your work process in the software.

Importering av CAD-modell, programmering enligt modell och ritning. Sedan används i princip samma programvara i maskinen där programmet körs.

Import of CAD model, offline programming according to the model and drawings. The same software is then used in the machines where the program runs.

How often do you perform new measurement preparations?

Vi gör en mätberedning per produkt. Kommer det en ny detalj som är snarlik så utgår vi från det befintliga programmet och ändrar. Men oftast skiljer de för mycket. Programmet körs från server och kan användas obegränsat antal gånger och i alla maskiner med Calypso.

We perform one measurement preparation for each product. Similar products starts from existing program with applied changes. But they usually differentiate too much. The program is run from a server and may be used any number of times and in all machines with CALYPSO preinstalled.


Appendix B:2

During our visit at your department you mentioned that the measuring points may be optimized whenever there is time for this. Is it that the time saved by an optimization isn´t that crucial for the overall measurement process?

Det finns mycket tid att spara. Inte så mycket gällande antal mätpunkter då detta optimeras från början, utan mer i mätordning och smarta funktioner. Detta blir det sällan tid till att föra in då det är mycket nyberedningar som måste fram.

There is a lot of time to save. The number of measuring points are already optimized from the beginning, but there is a lot to do in measurement orders and smart functions that may be applied. There is rarely enough time to do this, since there are many new preparations that must be handled.

You have mentioned earlier that the measurement department may become a bottleneck in the production; How often may this happen?

Alla detaljer skall igenom CMM, så det beror på inflödet. Är det jämt så flyter det på, men när det är ojämnt så blir det köer.

All components is measured in our CMMs in which depends on the inflow. If the inflow isn´t stable, there will be queues.

What do you think can be improved in your daily measurement preparation process?

Handlar främst om att få mer tid för att göra optimala program. Att få rätt indata i rätt tid

There is mainly to have more time to make optimal programs. Getting the right input at the right time.


PROJECT REPORT

Department of Engineering Science

C. Pre-study: The Basics of Geometry Assurance

January 05, 2015

The Basics of Geometry Assurance Robin Söderblom

Pre-study: The Basics of Geometry Assurance

i

The Basics of Geometry Assurance

Summary

Reducing costs and provide high quality products have been top prioritized among

manufacturing companies around the world for a long time. By implementing the geometry

assurance methodology in early product concept phase, unnecessary costs and rework due

to uncontrolled variations may be avoided. Geometrical variation effecting functionality and

esthetical aspects are critical quality characteristics that may originate from individual

manufacturing and assembly processes, must be controlled to assure a robust product.

Geometrical quality problems are mainly discovered in pre-production and before market

introduction, resulting in high costs for rework, market delays and bad publicity.

This report presents a methodology and approach of geometry assurance together with tools

that supports the robust geometry design in early concept phase, throughout the verification

and production phase. CAD systems are used to design virtual assembly models to be

analysed using CAT software where statistical analyses evaluate the robustness of a product

before realization. The methodology presented in this paper is gathered from different

scientific articles and PhD theses, mainly from the automotive industry, to be summarized

and explained in a holistic approach.

Date: January 05, 2015 Author: Robin Söderblom Examiner: Lecturer Mats Eriksson Advisor: Timo Kero, Semcon Sweden AB Main area: Mechanical Engineering Credits: 7,5 HE credits Keywords geometry assurance, GD&T, stability analysis, tolerance analysis, product

development Publisher: University West, Department of Engineering Science,

S-461 86 Trollhättan, SWEDEN Phone: + 46 520 22 30 00 Fax: + 46 520 22 32 99 Web: www.hv.se


ii

Preface

This paper is written as a report from a pre-study initiated at University West in cooperation

with Semcon Sweden AB to obtain knowledge and understanding in the field of geometry

assurance to be able to perform a good thesis work within the area. This pre-study has given

me good knowledge and understanding of the methodology and approach of robust

geometry design. I may now have the ability to work and practises the methodology and its

support tools in future assignments.

First of all I want to give many thanks to Timo Kero, Team Manager for the group Geometry

& Integration at Semcon Sweden AB, who is the sponsor of this project. Timo has shared

key contacts and materials that have given me a good start to the project.

Rikard Söderberg, Head of the department of Product and production development and Lars

Lindkvist, Docent/Associate Professor, Department of Product and Production

Development at Chalmers University of Technology, has given me and my work access to

the CAT software RD&T where I have practised and performed virtual simulations i.e.

stability analysis, tolerance analysis and contribution analysis. The software has given me

better understanding of the work with robust geometry design and I may now use this tool

in my upcoming thesis work.

Henrik Persson, Dimensional Systems Engineer at Semcon Sweden AB meet me over a

dinner to describe how their approach as an engineer and specialist of robust geometry design

are carried out to meet their customer requirements. He explained how the theories are

practiced in the reality and answered questions that the scientific articles don’t mentions.

Trollhättan, January 2015

_______________________________

Robin Söderblom


iii

Contents

Summary .............................................................................................................................................. i

Preface ................................................................................................................................................ ii

Symbols and glossary ....................................................................................................................... iii

1 Introduction ................................................................................................................................ 1 1.1 Geometry assurance ........................................................................................................ 1 1.2 Robust geometry design .................................................................................................. 2 1.3 The scope of the paper ................................................................................................... 2

2 The approach of geometry assurance...................................................................................... 3 2.1 Concept Phase .................................................................................................................. 3

2.1.1 Locating Schemes ............................................................................................... 3 2.1.2 Stability Analysis ................................................................................................. 7 2.1.3 Statistical Variation Simulation ....................................................................... 11 2.1.4 Tolerance Analysis ........................................................................................... 12 2.1.5 Tolerance Allocation ........................................................................................ 16

2.2 Verification Phase .......................................................................................................... 17 2.2.1 Inspection Preparation .................................................................................... 17 2.2.2 Locator Adjustments ....................................................................................... 18

2.3 Production Phase ........................................................................................................... 18 2.3.1 Root Cause Analysis ........................................................................................ 18 2.3.2 Six Sigma ............................................................................................................ 19

3 Conclusions and future work ................................................................................................. 20 3.1 Conclusions ..................................................................................................................... 20 3.2 Future work .................................................................................................................... 20

References ........................................................................................................................................ 21

Symbols and glossary

CAD – Computer Aided Design

CAT – Computer Aided Tolerancing

CTQ – Critical to Quality

GD&T – Geometry Design & Tolerancing

RD&T – Robust Design & Tolerancing

RSS – Root Sum Square

WC – Worst Case


1

1 Introduction

Manufacturing companies has always put a lot of effort to gain market shares through high

quality products by monitoring and controlling variation in the production. Magnusson,

Kroslid and Bergman [2] writes that 60-80% of quality problems are associated with errors

designed into the products and production processes during the design phase, see Figure 1.

Tolerances are introduced as a limit for variations to ensure good quality of components

through the entire process. Tight tolerances are very effective to assure fit and functionality

of their design but it becomes more difficult and expensive to produce. Therefore, apart

from the specific dimensions of a component, a complete system level specification must

also be considered to ensure the geometric tolerances, to account for the allowable ranges of

variation in geometry such that the full function is met at the minimum cost required.

Figure 1. Root causes for quality problems of products launched into the market. 63% of the cases originating from poor design according to Magnusson, Kroslid and Bergman [2].

1.1 Geometry assurance

A common issue in manufacturing today is dimensional variation stack ups where every

single component has its own dimension variation that occurs in an assembly. This can either

cause unexpected geometrical variation that propagates during the production and leads to

products that do not fulfil the esthetical, functional or assembly requirements, or the

tolerances for each component are unnecessary tightened. Those geometrical quality

problems contribute to high costs for rework, market delays and bad publicity due to changes

in product or production [1].

Geometry assurance is a methodology to manage variation and secure form, function and

assembly already in the concept phase by looking at the products tolerance chain and break

it down to component constraints and finally tolerances for individual geometrical features.

Marketing6% Purchasing

7%Quality4%

Production20%

Design63%


2

This is properly done by several activities throughout the product realization process shown

in Figure 2 [5].

Figure 2. The geometry assurance approach with different support activities throughout the product realization process, according to Söderberg, Lindkvist and Carlson [1]. These activities are described in section 2.

1.2 Robust geometry design

A robust product may be defined as a design that is insensitive against uncontrolled variation

or disturbance that may affect the performance, so called noise factors, generally originating

from manufacturing processes, temperature, wear, weather conditions and so on. A good

quality product may be characterize as a design that should be robust to noise. The activities

to improve robustness of a product are called robust design [5].

In robust geometry design, the main source of variation arises from the manufacturing

process. The main task is to optimize the location of locators in a way to minimize the

variation amplification to enable wider tolerances on input parameters to increase the

robustness of the design at a low cost [5].

1.3 The scope of the paper

In this work, a general approach to the geometry assurance methodology has been

summarized and presented in a way to get a central understanding of tools and methods used

in quality improvement for robust geometry design. Basic theories and examples are gathered

from related work and are presented in this paper to get a holistic approach for the

improvement process from an automotive industry perspective.


3

2 The approach of geometry assurance

This chapter describes the basics of geometry assurance and a set of tools integrated that

supports the quality improvement process. To enable a high impact from using the tools, the

work should begin in early product concept phase where only a few design characteristics

have been developed. Virtual parts and subassembly models are used to analyse the concept

design parameters (DP) to be evaluated against the functional requirements (FR)

continuously throughout the concept, verification and production phases. All virtual three-

dimensional (3D) models presented in this paper has been developed in a CAD software to

be analysed and improved using the CAT software RD&T. This software was initially

developed to evaluate variation within mass production for automotive and aerospace

industries. The tool are today used in various applications as an aid for geometric quality

improvements.

2.1 Concept Phase

In robust geometry design the main source of variation occurs mainly during the

manufacturing process. By increasing the robustness of a design in early concept phase, wider

tolerances may be used on input parameters which may result in decreasing manufacturing

costs. In this phase the virtual simulations is performed as early as possible and often starts

with a 3D CAD-shell model which is further developed through the design process. The

robustness of the concept design is optimized by locating systems and evaluated by stability

analyses. The esthetical quality level of the concept assembly is calculated and visualized by

statistical variation simulations which are verified against assumed production system by

tolerance analyses [1].

2.1.1 Locating Schemes

Locating schemes are used to position a part or sub-assembly in its correct position by

locking its six degrees of freedom in space during simulations, manufacture, assembly and

inspections. Locating schemes uses locating points called locators which are strategically

placed on a part or sub-assembly. These theoretical locating points are realized by physical

planes, holes and slots. The locators are extremely important since fixture tool variation will

be transmitted into parts and subassemblies which contributes to the robustness of the

design concept. The relation and dependencies can be formulated as equation (1).

[𝑟𝑜𝑏𝑢𝑠𝑡𝑛𝑒𝑠𝑠𝑣𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛

] = [𝑥 0𝑥 𝑥

] [ 𝑙𝑜𝑐𝑎𝑡𝑜𝑟𝑠

𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒𝑠 ] (1)

This equation indicates that the robustness is controlled only by the locators and should

therefore be focused on in early product and process concept phase. The main task in robust


4

geometry design is to place locators in a way to minimize the variation amplification [5]. This

locating procedure is supported by the stability analysis that will be described in chapter 2.1.2.

2.1.1.1 Choose locating scheme

There are different types of location schemes used in different situations. The most commonly used locating schemes are presented below and other less frequently used are only mentioned. 3-2-1 locating scheme for rigid parts Figure 3 illustrate an orthogonal 3-2-1 locating scheme with its six locating points. There are three groups of points called primary locating points, secondary locating points and tertiary locating point. These points are describe as follows:

- The primary locating points A1, A2 and A3, controls three degrees of freedom,

translated in Z (TZ) direction and rotation around X (RX) and Y (RY). These defines

the plane A.

- The secondary locating points, B1 and B2 controls two degrees of freedom,

translation in X (TX) and rotation around Z (RZ). These points defines the secondary

locating plane B, perpendicular to A.

- The tertiary locating point control one degree of freedom translation in Y (TY). This

point defines plane C, perpendicular to plane A and B.

In reality, the problem with all types of locating schemes is that they are coupled by nature.

This means that one locating point controls more than one degree of freedom. The 3-2-1-

system is the least coupled locating system since it enables the rotation and translation to be

minimized (decoupled). The ideal locating system is when one point only control one degree

of freedom. Figure 4 shows the main couplings for the 3-2-1 locating scheme [4].

This system is the most commonly used and is easier to use and understand among the

locating systems. It can be applied on non-prismatic parts that is assumed to be rigid [4].


5

Figure 3. 3-2-1- locating scheme are the most frequently used

locating system.

A1 A2 A3 B1 B2 C1

TZ

RX

RY

TY

RZ

TX

Figure 4. The coupling model for 3-2-1 locating system. The

columns represents the points inserted and the rows defines as

translated or rotation vectors. The grey area indicates the

couplings.

3-point locating scheme for rigid parts

Figure 5 illustrates an orthogonal 3-point locating system with primary locating plane A, secondary locating plane B and tertiary locating plane C. This system is a special case of the 3-2-1 locating scheme where three points are used as six points. The planes are described as follows:

- The primary locating plane A consists of A1, A2 and A3, locks three degrees of

freedom, TZ, RX and RY.

- The secondary locating plane B consists of A1 and A2, locks two degrees of

freedom, TX and RZ, perpendicular to A.

- The tertiary locating plane C, locks one degree of freedom TY, by A1,

perpendicular to A and B.

Due to its locating points, this locating system is commonly used for prismatic, rigid parts and the couplings are illustrated in Figure 6 [4].


6

Figure 5. 3-point locating scheme can be used for prismatic rigid

parts.

A1 A2 A3

TZ

RX

RY

TX

RZ

TY

Figure 6. The coupling model for 3-point locating system. The

columns represents the points inserted and the rows defines as

translated or rotation vectors. The grey area indicates the

couplings.

Other locating systems

There are other variants of locating schemes used for different types of robustness analysis.

The 3-2-1 and 3-point locating systems are used for orthogonal rigid parts only. When

working with non-rigid parts which are allowed to deform or bend during positioning such

as sheet metal or plastics, clamping forces and part stiffness has to be included to predict

robustness and variation. One example of locating system for non-rigid parts is the N-2-1

locating scheme, see [3].

Söderberg, Lindkvist and Carlson are proposing locating schemes for non-orthogonal

locating surfaces and locating systems [4].

2.1.1.2 P-frame

When working with positioning systems the usual notation P-frame is used for locating

schemes. Every part have one local P-frame in general, often referred to as the master

location system that positions the part to the mating target P-frame on another component

or subassembly [17]. See figure 7.

Figure 7. The 3-2-1 locating scheme referred as the local P-frame positioning to the mating target P-frame.


7


The stability analysis is a tool to analyse the geometrical robustness by evaluating the coupling

amplification and how much variation introduced to the component that is caused by the

locators [7]. To get a good understanding of the evaluation method, the theory of axiomatic

design will be explained as well as the theory behind the stability calculations. This will be

visualised by a simulation example to illustrate the software used.

2.1.2.1 Axiomatic Design

Axiomatic design is defined as the mapping process between customer needs trans-formed into functional requirements (FR), design parameters (DP) that physically satisfies the FR´s, and the process variables (aij) that represents the partial derivate aij=∂FRi/∂DPj at a specific design point [11] [17]. The process variables are included in the design matrix [A] also called the coupling matrix and may be written as equation (2).

[𝐴] = [

𝑎11 𝑎12 ⋯ 𝑎1𝑛

⋮ ⋱ ⋮𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛

] (2)

The design equation may be expressed as equation (3).

{𝐹𝑅} = [𝐴]{𝐷𝑃} (3)

By using the design equation in an example, the approach may be illustrated as equation (4) [17].

[𝐹𝑅1

𝐹𝑅2

𝐹𝑅3

] = [𝑎11 0 00 𝑎22 00 0 𝑎33

] [𝐷𝑃1

𝐷𝑃2

𝐷𝑃3

] (4)

Where there are three function requirements (FR) that may be satisfied by three design

parameters (DP). The left side of the equation, FR´s, may represents “what we want in term

of design goals” and the right side, [A] and DP´s, represents “how we hope to satisfy the

FR´s.” This design equation is the simplest case of design due to its non-diagonal elements

being zero, a12= a13= a21= a23= a31= a32= 0. This is characterized as an uncoupled design

which means that one output parameter is controlled by only one input parameter. The

diagonal characteristics is the most preferable design solution due to easier possibilities to

change FR´s or DP´s later in product or production phase. This situation often occur in

parallel assembly cases where all parts are attached to “ground” by its own P-frame and has

no influence from other parts or P-frames [17], see figure 8.

In serial assembly case, every part is attached to another part in a hierarchical order starting with part A, controlled by its own P-frame as an example. The following attachments B, C, D etcetera are controlled by its own P-frame and every P-frame mounted previously in the assembly as illustrated in Figure 9 [17]. This is characteristic for a decoupled design and may be written as equation (5).


8

[𝐹𝑅1

𝐹𝑅2

𝐹𝑅3

] = [𝑎11 0 0𝑎21 𝑎22 0𝑎31 𝑎32 𝑎33

] [𝐷𝑃1

𝐷𝑃2

𝐷𝑃3

] (5)

A decoupled design is an acceptable design if performed in the correct order to prevent time consuming tuning if necessary [17].

Figure 8. Parallel assembly model. Each P-frame controlling its

own P-frame only.

Figure 9. Serial assembly model. The last P-frame (part or

subassembly) attached is controlled by all P-frames.

A parallel assembly solution may therefore be preferred due to its low sensitivity to adjustments during manufacture and assembly compared to serial assembly solutions [17]. Look at a car door as an example, see Figure 10. The door may be positioned by two hinges and one locking mechanism on the car body. Every part have their own P-frame that is attached to the corresponding P-frame on the body. When looking at the couplings for the car door it may be written as equation (6).

[𝐻𝑖𝑛𝑔𝑒𝐴 𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛𝐻𝑖𝑛𝑔𝑒𝐵 𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛

𝐿𝑜𝑐𝑘 𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛

] = [

𝑎11 𝑎12 𝑎13

𝑎21 𝑎22 𝑎23

𝑎31 𝑎32 𝑎33

] [

𝑃 − 𝑓𝑟𝑎𝑚𝑒𝐻𝑖𝑛𝑔𝑒 𝐴

𝑃 − 𝑓𝑟𝑎𝑚𝑒𝐻𝑖𝑛𝑔𝑒 𝐵

𝑃 − 𝑓𝑟𝑎𝑚𝑒𝐿𝑜𝑐𝑘

] (6)

The positioning of the car door is highly depending on the P-frames for each of the parts involved. By changing one parameter, all other P-frames are affected and may result in time consuming production set-up. The concept becomes sensitive to disturbance and the mounting system introduced to attach the door sub-assembly to the body makes the design sensitive to variation. This is called a coupled design which is preferred to reduce or eliminate. A big challenge within geometry assurance is to design concept solutions that avoids non-diagonal couplings to prevent expensive costs for changes in product and production development.

2.1.2.2 Robustness evaluation

The variation for a sub-assembly or assembly is calculated by varying each locating point, i

for a part with a small increment Δinput. Δoutput/Δinput can be determined in the X,Y and

Z directions for a number of n output points of the geometry [5]. The root mean square

(RMS) values for all output points corresponding to variation in the locating points, are then

calculated in equation (7)(8)(9).


9

𝑅𝑀𝑆𝑥,𝑖 = √1

𝑛∑ [

(𝑥−𝑥𝑛𝑜𝑚)

Δinput]

2𝑛1 (7)

𝑅𝑀𝑆𝑦,𝑖 = √1

𝑛∑ [

(𝑦−𝑦𝑛𝑜𝑚)

Δinput]

2𝑛1 (8)

𝑅𝑀𝑆𝑧,𝑖 = √1

𝑛∑ [

(𝑧−𝑧𝑛𝑜𝑚)

Δinput]

2𝑛1 (9)

The resulted RMS values represents the mean influence of all locating points, i, are calculated

in each direction separately to evaluate the total positioning evaluation goodness for the total

Root Sum Square (RSS) magnitude [5]. The RSS influence in all locating points, i, is calculated

in each direction as well in equation (10)(11)(12).

𝑅𝑆𝑆𝑥 = √∑ 𝑅𝑀𝑆𝑥,𝑖26

𝑖=1 (10)

𝑅𝑆𝑆𝑦 = √∑ 𝑅𝑀𝑆𝑦,𝑖26

𝑖=1 (11)

𝑅𝑆𝑆𝑧 = √∑ 𝑅𝑀𝑆𝑧,𝑖26

𝑖=1 (12)

𝑅𝑆𝑆𝑥,𝑦,𝑧 = √𝑅𝑀𝑆𝑥2 + 𝑅𝑀𝑆𝑦

2 + 𝑅𝑀𝑆𝑧2 (13)

The RSS magnitude, RSSx,y,z in equation (13) used as a sensitivity value to evaluate how a

certain P-frame controls the stability of a certain part in a design [5].

To evaluate the coupling dependencies for an assembly, two measures are introduced as

reangularity R and semangularity S. Read more about those in Suh [6]. These values represent

the diagonality of the design matrix and are defined as equation (14)(15) [5] [17].

𝑅 = ∏ [1 −(∑ 𝑎𝑘𝑖𝑎𝑘𝑗

𝑛𝑘=1 )2

(∑ 𝑎𝑘𝑖2𝑛

𝑘=1 )(∑ 𝑎𝑘𝑗2𝑛

𝑘=1 )]

1/2

𝑖=1,𝑛−1𝑗=1+𝑖,𝑛

(14)

𝑆 = ∏ [|𝑎𝑗𝑗|

(∑ 𝑎𝑘𝑗2𝑛

𝑘=1 )1/2]𝑛𝑗=1 (15)

aij is an element from the design matrix described earlier, and n is the number of rows in the

design equation. If R=S=1, the design matrix represents an uncoupled design which is the

ideal. R and S provides good possibilities to evaluate the degree of coupling in an early

concept assembly to avoid unnecessary costs for changes in production ramp-up. Concepts


10

requiring tight tolerances due to tolerance chains may be sorted out in an early stage using

the stability analysis [17].

2.1.2.3 Using Computer Aided Tolerancing software

The evaluation method requires computer aided virtual simulations to efficiently calculate

the variation amplification for complex systems. Since the relation between geometrical

features of parts and sub-assemblies are constrained by tolerances, the final variation of the

resulting assembly is often controlled by the individual component dimensions. The only

way to change the variation of an assembly is to change the allowed variation of individual

part features which is a result of tolerance chains i.e. geometrical couplings [5].

An example is used to show how the robustness of a design may be increased to easier

allocate and modify critical tolerances by changing the way individual components are located

and assembled. Figure 10 shows a stability analysis of a door mounted on an automotive

body, as mentioned in section 2.1.2.1. The design example illustrates in what way the door is

positioned and assembled on the body by two hinges mounted to the door by its

corresponding P-frames. The door and the two hinges creates a sub-assembly which is then

mounted to the body with a compound P-frame consisting of locating features such as

planes, holes and slots from the hinges, and the door represented by the actually locking

system. This results in a very sensitive design with high couplings framed with red in the

figure. A change in one feature may affect the position of all other parts in the subassembly.

Adjustments of this design in product realization phase may cause high cost for time

consuming changes and are very sensitive for disturbance during the production [5].

Figure 10. A sensitivity analysis of a door sub-assembly mounted on an automotive body. The red areas in the table represents coupling amplifications in the design.

By introducing a fixture system, the overall robustness for the subassembly may be improved.

The solution locates the door in the correct position with a fixture and the hinges are


11

positioned directly to the body. This results in a broken tolerance chain with less couplings

involved. Figure 11 shows a version of a fixture where the door is only controlled by its own

locating system and the fixture P-frame. The hinges are here mounted to the body with their

corresponding locating systems controlled entirely by its own P-frame and the body P-frame.

This solution improves the R and S- values in the analysis. A change, for example, in one

hinge will not influence any other part positions in this design. Costs for fixturing and related

quality level have to be compared with the need and costs for longer tolerance stackups with

tighter tolerances as a result [5].

Figure 11. A stability analysis performed with a fixture system introduced to cut the tolerance chain. The couplings over the diagonal are eliminated and there are only a few dependencies below the diagonal.


To be able to determine the quality level for a product before production realization, the

geometrical variation for critical features must be predicted. By performing a statistical

variation simulation on critical assembly dimensions, the product may be analysed and

improved if necessary before the first prototype is build. By using the Monte Carlo method

(briefly described in section 2.1.4.1), the software randomly generates all input parameters

within the defined distributions for every part and simulates the resulting output parameters.

For a number of iterations the simulation predicts the expected output standard deviation,

range, mean value and capability indices etcetera for the specified dimension to determine if

the critical feature meets the requirements [1]. Figure 12 illustrates a Monte Carlo simulation

for a critical part dimension on a Volvo S80 where the flush between the boot lid and the

left rear lamp are analysed. The critical dimension is here defined as a point-to-point on each

part together with a measuring direction.


12

Figure 12. Statistical variation simulation using the Monte Carlo method to predict the output standard deviation, range, mean value and capability indices. Here, 1000 iterations was performed to analyse the flush between the boot lid (blue area) and the left rear lamp (red area).

Statistical variation simulations can also be performed on non-rigid parts like plastic materials and sheet metal. By integrating FEA techniques, over-constrained P-frames with more than six degrees of freedom may be analysed to see how the parts behave after assembly [13]. Seam variation analysis

The relation between two parts over a specified distance in assembled products, describes

the most frequently used quality characteristics for geometrical variation evaluation,

especially in automotive body design. The quality of the split-line between two body panels

of a vehicle, for example the doors, hoods and panels are critical quality characteristics due

to its functional and esthetical aspects. The door must be possible to open without any

conflict with surrounding parts while the split-line, mainly translated by the gap, flush and

parallelism, must satisfy its desired quality requirements [18].

The gap represents the distance between two parts in a common plane, See figure 13, while the flush refers to the distance between two parts perpendicular to a surface or a plane. These dimensions are often measured or calculated for two specific points, where each point are located on each part in a specified 2D-plane [7].

Figure 13. Seam variation illustrated by gap and flush.

2.1.4 Tolerance Analysis

In order to control if the resultant variation for a part or assembly, caused by tolerancing,

meets the FR´s, a tolerance analysis is commonly used. The purpose of using tolerance


13

analysis is to determine the effect of variations caused by each specified tolerance, called the

contributor. All the known tolerances that influences the total variation of a dimension

contributes to a tolerance chain, called stackup, see figure 14. This analysis is thereby known

as the stackup analysis or design assurance [8].

Figure 14. Variation caused by a few individual tolerances in a stackup may result in a massive resultant.

A tolerance chain arises when a critical component dimension is dependent of another

individual dimension. Figure 15 illustrates a simple one-dimensional (1D) tolerance chain

occurring on a vehicle floor consisting of a tunnel for the cardan shaft and two separate floor

panels on both sides. The resulting width of the floor is controlled by the sum of every

component dimension. The overall variation is thereby controlled by every individual part.

This makes the design very sensitive and results in difficulties to adjust the floor width during

assembly. By using this solution, the manufacturing process may be more expensive to assure

high quality due to tighter tolerances for every component [11].

Figure 15. The total floor with is controlled by each individual part dimension.

The most efficient solution from an economic perspective is to eliminate all stackus by

redesigning the concept. Figure 16 illustrates a different solution to the stackup problem. By

allow the parts overlapping each other and including a fixture to adjust the floor width during

assembly, the solution become more robust and minimizes the high demands on tight

tolerances for every part. The fixtures on both sides controls the overall width and the fixture

pin controls the tunnel position which may be adjusted just before production starts. These

fixtures are used in the assembly process and are being removed later [11].

Figure 16. Components are overlapping to avoid stackup dependencies and fixtures are used to control the total floor width.


14

If robust design and ease of adjustments to a specific quality level during production shall be obtained, tolerance chains must be avoided and this is one example to an alternative solution to eliminate a critical tolerance chain.

2.1.4.1 Stack-up models

When working with complex three-dimensional (3D) assembly design, tolerance chains may

be difficult to identify and handle in production. Tolerance analysis is therefore preferably

performed before the final geometry is set to detect potential tolerance stackups and increase

geometrical robustness.

There have been several methods for performing tolerance analysis for rigid components

developed through the years. The worst-case (WC) model, Monte Carlo and a number of

statistical models presented in figure 17 are variants used for the analysis [12].

Figure 17. Different mathematical models for tolerance analysis presented in Chase [12].

Mean shift and Six sigma are variants of Root sum square.

The most widely used statistical models is the Root Sum Square (RSS) which will be described

and compared with two other frequently used models, namely the worst-case and the Monte

Carlo model [12].

Worst Case

The worst-case model is the simplest model among the others presented in Figure 17 and is

based on the arithmetic law. It assumes that all tolerances in a stackup are at its extreme limit

simultaneously to obtain the worst possible combination of parts. If the WC stays within the

required tolerance limits, there no rejected assemblies needed. The stack formula is non-

statistical and may be written as equation (16) [1] [12].

𝑑𝑈 = ∑|𝑇𝑖| ≤ 𝑇𝐴𝑆𝑀 (16)


15

Where dU is the predicted assembly variation, Ti the component

tolerance allowed for one specific dimension and TASM is the

maximum respectively minimum tolerance limit required for the

overall assembly.

This method is commonly used by designers in early concept stages

to assure that their assemblies stays within the specified tolerance

limits, but is also preferably performed during manufacturing where

the production volume is low and the tolerance chain is short [1].

Statistical RSS

By adding variations to the calculation, the predicted limits are more

reasonable due to its statistical probabilities of the possible

combinations. The RSS model is the most simple of the statistical

models and assumes a normal distribution of component tolerances

in an assembly. This model is applicable in high production volume

and longer tolerance chains but may have an optimistic result. The

predicted assembly variation may be written as equation (17) [1] [12].

𝑑𝑈 = √∑ 𝑇𝑖 2 ≤ 𝑇𝐴𝑆𝑀

(17)

As an example, figure 18 illustrates an assembly stackup with eight boxes of equal precision due to same tolerance Ti = 2 mm. The predicted total variation for the assembly height by using WC would result in equation (18).

𝑇𝐴𝑆𝑀 = ∑|𝑇𝑖| = 8 × 2 = ±16 (18)

The result from RSS would be equation (19).

𝑇𝐴𝑆𝑀 = √∑ 𝑇𝑖 2 = √8 × 2 2 = ±5,66 (19)

WC predicts much more variation than RSS and the differences becomes more significant as

the number of component dimensions in the stackup increases.

By doing the opposite, the component tolerances are determined from the assembly

tolerance requirement instead. Suppose that the TASM = 16 is specified as the design limit for

the assembly, WC results in equation (20).

𝑇𝑖 =𝑇𝐴𝑆𝑀

8=

16

8= ±2 (20)


16

RSS results in equation (21).

𝑇𝑖 =𝑇𝐴𝑆𝑀

√8=

16

√8= ±5,66 (21)

WC requires tighter tolerances for each component than RSS to meet

the same assembly requirements. RSS is here preferred from an

economic perspective due to the resulted variation allowed.

The equations presented for WC and RSS only illustrates the basic

theory for the models and these becomes more complex in real assembly systems. Kenneth

, Chase and Parkinson [9] writes about factors such as sensitivities, correction factors and

mean shift factors that are introduced in the calculation to get a more realistic prediction.

Monte Carlo Samples

Monte Carlo simulations have been a very frequently used tool for tolerance analyses since

it handles both non-normal distributions and nonlinear assembly functions [1]. It uses a

number generator to randomly simulate the effect of variation from manu-facturing

processes by generating output parameters for every dimension of an assembly and continues

iteratively until a sufficient number of iterations has been simulated to plot a histogram. A

percentage of rejected assemblies is estimated based on the specified tolerances.

This method is sample-based but the simulation becomes statistical with enough number of

samples and require more than 100 times the CPU capacity compared to WC and RSS [9].

The Monte Carlo simulation is very useful in 3D tolerance analysis and is used as basis for

the majority of today´s CAT systems [1].

2.1.5 Tolerance Allocation

Optimizing performance, quality and production costs often requires tolerance allocation to

strategic allocate the critical contributors. Tolerance allocation is often described as the

opposite of tolerance analysis due to the purpose to break down tolerance chains in an

assembly to locate the individual contributors [1]. By performing a contribution analysis, the

contribution for each part variation is calculated and the most critical tolerances may be

prioritized and investigated. The basic formula for each part variation may be written as

equation (22) and (23) [12].

𝑊𝐶: %𝐶𝑜𝑛𝑡 = 100𝑇𝑖

𝑇𝐴𝑆𝑀 (22)

𝑅𝑆𝑆: %𝐶𝑜𝑛𝑡 = 100𝑇𝑖

2

𝑇𝐴𝑆𝑀2 (23)

Figure 18. Stackup illustration with boxes. Units in mm.


17

Figure 19 shows a contribution analysis performed on a gap variation between the left lamp

and the trunk lid for a Volvo S80. The analysis resulted in a number of contributors and the

top five contributes the most to the gap variation. The first contributor (1) is located as a

point at the trunk lid defined with a tolerance range of 2 mm and contributes 20,1% of the

overall gap variation in the area. The second contributor (2) is a point located on the left

lamp with a tolerance range of 2 mm and contributes 20,1%. The third contributor (3) is a

point located inside the left lamp and represents a bolt joint that positions the lamp in one

direction with a tolerance of 1,5 mm and contributes 12,8%. The fourth contributor (4) is

located inside the body to position the trunk cover with a bolt joint and a defined tolerance

range of 1,41 mm which contributes 10%. The fifth contributor (5) represents another bolt

joint that attaches the left lamp to the body with a tolerance of 1,5 mm which contributes

9%. By changing the tolerance range for the first contributor in the example, the output

variation for the gap between the left lamp and the trunk lid may have a significant

consequence.

Figure 19. A contribution analysis was performed to investigate what contributors that contributes the most to the gap variation between the left lamp and the trunk lid at a specific point.

2.2 Verification Phase

In the verification and pre-production phase the virtual product models are physically

realized to be tested and verified against the production system. Here, adjustments are made

for both the product and the production system to prepare for full production volume. In

geometry design the verification phase involves inspection preparation where inspection

routines and strategies are established. The virtual assembly model is used to minimize the

geometry errors by locator adjustments and to support the inspection plan [1].


The aim of inspection preparation is to find the optimum set of inspection points that

captures product information to verify if adjustment, correction or compensation is


18

necessary. Often the number of inspection points becomes quite large in pre-production to

access a large quantity of process information to monitor the actual process level. During

full production the number of inspection points are reduced to only capture key dimensions

on every product and the measuring process is carrying out in every assembly stages to

monitor the quality level through the entire assembly process. Söderberg, Lindkvist and

Carlson [1] illustrates how this activity may be performed using virtual variation simulation

models to support the inspection preparation.

2.2.2 Locator Adjustments

Form errors are often discovered during the pre-production which may cause functional or

esthetical complications. Often the solution have been to compensate the problem by

reposition the involved components by modifying their locators. This activity is known as

trimming and is traditionally done by assembling a number of parts, checking the deviations

to surrounding components, disassembly the components, adjusting the positioning points,

reassembling the parts again and measure the new result. This procedure may be time and

effort consuming due to repeated operations until a satisfactory result is achieved [1].

Söderberg, Lindkvist and Carlson [1] mention an alternative solution called virtual trimming

where the adjustments is done using a virtual computer tool based on inspection data from

the initial components in association with variation simulation models. The tool simulates

the result directly and includes optimization of the trimming, which eliminates the time

consuming physical tests.

2.3 Production Phase

During the production phase all adjustments from the pre-production are completed, the

product geometry design satisfies the FR´s, the geometrical tolerances may be accepted and

the product is in full production. Samples may be analysed during the production process

for distinguishing between the common causes due to noise factors and assignable causes

often due to defect raw material, operator errors, machine and tooling errors. These

variations may be controlled and minimized or eliminated by Root Cause Analyses (RCA) or

by the Six sigma (6σ) approach [1].

2.3.1 Root Cause Analysis

The objective of RCA is to efficiently trouble shooting the dimensional error in an assembly

plant by pinpointing at the root of variation. The effort of eliminating or limiting these

sources of variations contributes to the long term process improvements. Complex products

such as an automotive body are assembled in a multilevel hierarchy by both serial and parallel

subprocesses using different assembly tools. During the production process, geometrical

errors may propagate from a variety of factors, see Figure 20, and is likely to be difficult to

identify [1]. Carlson and Söderberg [14] presents a computer tool for RCA that allows

individual locator, fixture and station errors to be identified. The tool enables adjustments in


19

process parameters by translating variation and deviation into complex geometry data.

During the production process the virtual simulation model is fed with inspection data from

the contemporary production to be analysed if the occurred product error originate from the

assembly fixtures which may be located to a specific locator.

Figure 20. Cause and effect diagram of assignable causes that may contribute to the overall assembly variation [14].

2.3.2 Six Sigma

Six Sigma is a company-wide improvement methodology for controlling processes with aim

to reduce cost and increase revenue. It was pioneered by Motorola in 1987 as a strategic

initiative and has become one of the most used improvement methodology for large, medium

and small sized enterprises worldwide. This methodology is a project by itself and consists

of five operations defined by DMAIC [1] [2].

It starts with the define phase where the product or process that need improvement is

identified and specified in a project charter.

In the measure phase, the mainly activity is to collecting performance data on critical-

to-quality (CTQ) by selecting (Y´s) of a product or process that are to be improved,

as well as their input parameters (x´s), that influence (Y). Data from input and output

parameters are then selected.

The following analyse phase is performed by analysing and calculating the mean value

and variation from data gathered for input and output to identify critical sources of

variation for CTQs.

The focus in the improvement phase is to find the input (x´s) which contribute the most

to the output and find the best solution by shifting their mean and/or reduce their

variation or influence. This can be done by performing a contribution analysis on the

variation simulation model considers both geometrical sensitivity and variation

amplitude.

When the solution has been implemented the (Y´s) are monitored during the control

phase to verify that the requirements has been achieved.


20

3 Conclusions and future work

This chapter contains conclusions of the information presented in this paper and discusses

proposals for future work.

3.1 Conclusions

The methodology strives for strengthening the attractiveness of a product in order to

improve the overall quality, which in turn, increases the customer´s impression of product

quality. This is especially applied in the automotive industry where many brands and more

to come strive to reach premium class quality. Perceived quality has always been associated

with premium products which motivates the industry to focus on the customer´s perception

of perceived quality. The tools, methods and knowledge presented in this paper is an

important aid for early concept evaluation and improvement to achieve customer

requirements and perceptions of high quality.

This paper has been compiled as a short summary from scientific articles and PhD theses

within the area of geometry assurance methodology. The approach presented in section two

are implemented in the automotive industry and examples illustrated are cases that may arise

in the field of quality work. A lot of information has been gathered from scientific articles

written by Swedish scientists where methods and tools are implemented at a Swedish car

manufacturer. The information presented may be interpreted as a general approach within

the automotive industry but may vary depending on different cases.

The activities presented in the concept phase are here explained in more detail than the other

activities mentioned in the verification and production phase. This is due to limitations in

the work of this paper to focus on the main activities to improve the product quality in early

concept phase to avoid unnecessary costs.

3.2 Future work

The tools and methods presented in this paper should provide a basic understanding of the

improvement work within the geometry assurance methodology. With this knowledge, you

should be able to exercise stability analyses, tolerance analyses and contribution analyses of

simpler systems as well as understand and practice the geometry assurance methodology

from early concept phase, throughout the verification and production phase.


21

References

2. Söderberg, Rikard., Lindkvist, Lars & Carlson, Johan (2006). Virtual geometry assurance for

effective product realization. 1st Nordic Conference on Product Lifecycle Management –

NordPLM´06 25-26 January, 2006, Göteborg.

3. Magnusson, Kjell, Kroslid, Dag, Bergman, Bo, Häyhänen, Peter & Mills, Don

(2003). Six sigma: the pragmatic approach. 2. ed. Lund: Studentlitteratur

4. Cai W., Hu S. J., Yuan J. X. Deformable Sheet Metal Fixturing: Principles, Algorithms,

and Simulations, J. Manuf. Sci. Eng. 118, 318-324 (1996).

5. Söderberg, R, Lindkvist, L, & Carlson, J 2006, 'Managing physical dependencies

through location system design', Journal Of Engineering Design, 17, 4, pp. 325-346.

6. Söderberg, R, Lindkvist, L, & Dahlström, S 2006, 'Computer-aided robustness analysis

for compliant assemblies', Journal Of Engineering Design, 17, 5, pp. 411-428.

7. Suh, N.P. (1990) The principles of design. New York: Oxford University Press

8. Söderberg, R, Lindkvist, L 2002, Stability and seam variation analysis for automotive

body design, Journal Of Engineering Design, 13, 2, pp. 173-187.

9. Wagersten, Ola (2013). Visualizing the Effects of Geometrical Variation on Perceived

Quality in Early Phases. Diss, Gothenburg: Chalmers University of Technology.

10. Kenneth, W, Chase & Parkinson, A.R 1991, 'A survey of research in the application of

tolerance analysis to the design of mechanical assemblies', Research in Engineering Design,

3, 1, pp. 23-37.

11. Shen, Z, Shah, J.J. & Davidson, J.K. 2008, 'Automatic generation of min/max tolerance

charts for tolerance analysis from CAD models', International Journal of Computer Integrated

Manufacturing, 21, 8, pp. 869-884.

12. Söderberg, R, & Johannesson, H 1999, 'Tolerance Chain Detection by Geometrical

Constraint Based Coupling Analysis', Journal Of Engineering Design, 10, 1, pp. 5-24.

13. Chase, K. (2004). Basic tools for tolerance analysis of mechanical assemblies.

Manufacturing engineering handbook. Utah: Brigham Young University.

14. Wagersten, O, Wickman, C, Lindkvist, L & Söderberg, R 2013, 'Towards non-FEA-

based deformation methods for evaluating perceived quality of split-lines', Journal Of

Engineering Design, 24, 9, pp.623-639.

15. Carlson, J.S, Söderberg, R 2003, 'Assembly Root Cause Analysis: A Way to Reduce

Dimensional Variation in Assembled Products', The International Journal of Flexible

Manufacturing Systems, 15, 2, pp. 113-150.


22

16. Ulrich, Karl T & Eppinger, Steven D (2012), Product Design and Development, 5th ed.,

ISBN 978-007-108695-0 (Paperback), McGraw-Hill

17. Bergman & Klefsjö (2010), 'Quality from Customer Needs to Customer Satisfaction.,

Studentlitteratur, Sweden, 3:rd edition, 2010.

18. Lindkvist, L & Söderberg, R 1999, 'Computer Aided Robustness Evaluation', Journal of

Engineering Design, pp. 165-181.

19. Dagman, A, Söderberg, R, Lindkvist, L 2007, 'Split-line design for given geometry an

location schemes', Journal of Engineering Design, pp. 373-388.

Instruction Manual: The process to generate CMM programs with RD&T and IPS

Appendix

D. Instruction Manual

Instruction Manual: The process to generate CMM programs with RD&T and IPS

Appendix